Luận văn thạc sĩ Quản lý xây dựng: Analyze the impact of factors affecting rework costs in construction projects using ann

18 Table 2.5 Ranking the causes and impact of rework factors related to the design according to Al-Janabi, 2020 .... 19 Table 2.7 Ranking the causes and impact of rework factors relate

INTRODUCTION

Research Statement

In the construction field of the world in general and in Vietnam, the quality of each project is considered an important factor compared to progress and costs during the completion phase Project quality is the aspect to show the success and trust from users Due to the importance of quality, it is required to clearly identify the tasks of all entities involved in the process of creating construction products, including Investors, Consultants, Contractors and related stakeholders in design, construction, management and operation construction project In other to have a good quality construction project, it is necessary to ensure that the work in advance and during the construction phase must be good such as design, construction work and external factors However, in reality, design and construction work often has to be repeated many times, also known as "rework" for many different reasons during the construction process Furthermore, such rework may result in overtime, additional hiring of resources, schedule slippage, or reductions in project scope or quality The consequences of these difficulties include reduced profit, loss of market share and reputation, increased turnover of management and workforce, lower productivity, higher costs, and all too frequently costly litigation between participants over responsibility for overruns and delays according to Love, P E D., et al

Rework is a common issue in construction and engineering projects, affecting performance and cost growth On average, rework contributes to 52% of total cost growth and can increase schedule overrun by 22% Costs range from 5% to 20% of contract value, with changes in design scope accounting for up to 50% of rework Rework costs can range from 5% to 20% of the contract value The primary objective of the research reported in this paper was to determine the influence of project and managerial aspects on rework costs, using data from incidents that occurred at construction sites In design and construction projects, design flaws that result in rework and/or design revisions are thought to be the main cause of schedule delays and cost overruns Even though design faults are thought to be common, the majority of design and construction firms do not track the quantity of errors they produce, thus they do not fully understand how these the dynamics of design flaws and methodically evaluate their detrimental effects in order to solve this The model's creation and application to a university building project are covered in this study The findings show that even with construction managers' constant efforts to recover the timeline, design flaws can cause a project's timetable to be severely delayed The study confirm to the created model's ability to more thoroughly evaluate the detrimental effects of design flaws, which practitioners frequently undervalue These findings lead to the conclusion that, especially when time is of the essence, the created model can help project managers gain a better understanding of the dynamics of design flaws and recover delayed schedules according to Han, Love et al 2013 [2]

Reworks in construction projects significantly impact project outcomes, increasing costs and time, and requiring additional resource use The construction sector consumes significant natural resources and generates significant pollution due to the reworking process, which also involves the use of electricity, fuels However, one must not undervalue the quantity of resources including energy that are used during the building stage The issue of reworks in building projects takes on significance in this setting In addition to the evident detrimental effect on the project outcome, manifested in higher expenses and duration, the reworks result in a greater consumption of resources following to Trach et al 2021 [3] Moreover, For the majority of nations, construction is one of the most significant sectors, and the success of the economy as a whole depends on it Construction generates a lot of employment and consumes intermediate goods and services (up to 40% of raw materials, chemical products, electrical and electronic equipment, etc.), making it a vital sector of the national economy The building sector's performance has a big impact on the growth of the economy as a whole The manufacturing of building supplies and equipment, mechanical engineering, metallurgy and metalworking, petrochemicals, glass manufacturing, woodworking, transportation, and energy all expand with the construction industry as mentioned by Trach, R., et al

In Vietnam, cost overrun have been known as the top reason leading to extend construction schedule in long period and become bad face for society So far, there have been some researches about overcost within construction stage at projects in Vietnam such as Le-Hoai, Long & Lee (2008) [5], Tran Pham Khanh Toan (2021) [6] and Tran Hoang

Tuan (2014) [7], especially the research for “rework” factors affection on Design stage of Truong and Cao (2021) [8] However, detail research about factors which affect to rework cost in Vietnam projects have been published widely

As explanation above, this research article have been identified to discuss and determine main factors impact on rework cost in construction project and then using AI method to analyse impact level of each factor.

Research Problem

Research problem is determined factors which frequently occur and impact level of factors on Rework cost on contruction projects in Vietnam The question would be raised that:

 What are factors impact on rework cost on construction projects?

 Which would factors be evaluated with the most affection on rework cost?

 How will construction parties react to reduce and minimize rework factors?

Research Objectives

 Determine factors impact on rework cost from aspects such as Investor, consultant, contractor, external factors, labour and equipment in construction project in Vietnam

 Apply SPSS for analyzing and ranking factors with the most impact on rework cost

 Apply ANN to analysis rework cost impact Then comparing ANN model with other model for discussion.

Scope of Study

 Survey location: Projects all over Vietnam, especially in the south of Vietnam such as Ho Chi Minh City, Dong Nai, Binh Duong, Long An, Ba Ria- Vung Tau Province,

 Survey project: All projects have been completed from 2013- 2023

 Survey object: Engineer and employee in construction field with experience in construction project management from investor, consultant (design and supervision), contractor, subcontractor, supplier, other parties related to construction field

 Survey period: From February 2024 to May 2024

Contribution to Academic and Practical Fields

 Introduce AI Application method ANN to analysis and evaluate factors which special impact on rework cost in current construction project

 Compare ANN method with previous methods in order to get accurate identify of rework costs in construction

 The study shows the potential of this method for analyzing factors that greatly impact rework costs of construction projects

 The research can be applied to departments that manage construction projects strategically to limit the impacts that lead to reconstruction cost

 Refer and choose appropriate management methods to bring efficiency in construction cost and time.

Outline of thesis

Thesis would be included 6 chapters, include:

LITERATURE REVIEW

Concepts and Definitions

According to Construction Law 62/2020/QH14 Article 3 Section 10 [9]: A construction project is a product built according to design, eracted by human labour, construction materials, and equipment installed in the project, linked to the land, and may include parts underground, upper ground, under water and above water Construction projects include public, civil project, residential project, industrial project, transportation project, irrigation project, energy and other projects

Figure 2.1 Type of construction projets in Vietnam

 Investor: is the party who owns the capital or is the party assigned to manage and use capital to invest in the construction project

 Consultant: is the party who are signed contract with Investor for coordination to manage and support about construction works which Investor is not enough capacity to implement It is considered many type of

ProjectIndustrial Project consultants in construction work such as Design consultant, supervision consultant, bidding consultant, project management consultant, Design consultant and Supervison consultant are common with all of construction projects

 Design consultant: Design consultant is a professional design company in which experts with knowledge and experience in the field of construction provide design, planning and construction project management solutions to investor This service includes supporting investor from the initial concept stage through to project completion Design consultant is not only helps investors optimize design and costs but also ensures technical standards, safety, legal compliance and aesthetics

 Supervision consultant: is the party sign contract with Investor and cooperate with Project management unit to supervise construction works in project Moreover, supervision consultant support investor control and manage all of construction items to ensure about quality and construction schedule

 General Contractor: A general contractor is a contractor who signs a contract directly with the investor to undertake all of a type of work or all of the work of a construction investment project General contractors include the following main types: general design contractors; general contractor for construction works; general contractor for design and construction of projects; General contractor for design, supply of technological equipment and construction of projects;

 Main Contractor: The main contractor is the contractor who signs a contract directly with the construction investor to construct the main part of a type of work of a construction project

 Subcontractor: A subcontractor is a contractor who signs a contract with the main contractor or general construction contractor to carry out part of the main contractor's or general construction contractor's work

 Supplier: is party signed contract with main contractor/ Subcontractor or general contractor for supplying materials which are used during construction period

Figure 2.2 Relevant of construction project

2.1.3 Total investment amount for construction project

According to Decree 10/2021/NĐ-CP about Construction investment cost management [10], total investment amount include: compensation, support and resettlement costs (if any); construction cost; equipment costs; project management cost; construction investment consulting cost; other costs and Contingency cost

 Construction cost: include construction work costs and project items; temporary and auxiliary construction works and items serving construction; Demolition cost for construction works that are not within the scope of site clearance demolition work have been determined in compensation, support and resettlement cost

 Project management cost: are the costs necessary to organize and manage the construction activities and implementation of project management tasks from the preparation stage, construction stage to the completion of construction project into operation and settlement of construction investment capital

 Equipment cost: are the costs of construction and technological equipment procurement; equipment procurement management costs (if any); Cost of purchasing software licenses used for construction equipment and

Subcontractor Supplier any); costs of processing and manufacturing equipment that needs to be processed and manufactured (if any); installation, testing and adjustment costs; Cost of testing equipment according to technical requirements (if any); transportation costs; insurance; taxes and fees; other related costs

 Compensation, support and resettlement costs: include compensation costs for land, houses, structures on land, assets attached to land, on water and other compensation costs according to regulations; subsidies when the state recovers land; resettlement costs; costs of organizing compensation, support and resettlement; land use and land rental costs calculated during the construction period (if any); relocation costs, reimbursement for technical infrastructure that has been invested in construction to serve site clearance (if any) and other related costs

 Construction investment consulting costs: are the costs necessary to carry out construction investment consulting work from the project preparation stage, project implementation to the end of construction and when construction project come into operation, using and settlement of construction investment capital

 Other cost: include necessary costs to implement construction investment project such as mine and explosive material clearance; Cost of returning technical infrastructure due to impacts during construction; material storage costs; construction insurance costs during construction; initial working capital for construction investment projects for business purposes, loan interest during the construction period; construction deformation monitoring,

 Contingency cost: include contingency costs for additional work and quantiy, contingency costs for price inflation factors during project implementation

2.1.4 Cost overrun construction project definition:

Cost overrun of a construction project is when the final cost are over with estimation budget and which is incurred by many factors during the construction process such as arising scope of work, unexpected events, other costs to rework during the construction process, external factors affecting material costs, etc However, currently very few projects are completed without cost overruns, which is a problem that investors, consultants and contractors most interested in making decisions Cost overrun cause many affections about construction schedule which leads to serious problems in society

2.1.5 Cost overrun construction project type:

 Cost overrun by design: caused by lack of design documents such as details, regulation, material information, poor coordination between design parties (Architecture, Structure, MEP)

 Cost overrun by management: caused by lack of management skills and experience, poor communication between parties, lack of project overview

 Cost overrun by consulting: caused by lack of construction experience, construction method omission, lack of understanding for construction law, TCVN regulation, construction standard

 Cost overrun by construction: caused by lack of construction skills, poor quality material using, lack of following design information, rework items due to wrong or missing construction work

Rework is a non-value-adding process in construction projects, causing a discrepancy between the actual and planned state of an object and requiring compliance to improve productivity and performance [3] A term defination of rework that is frequently used but quite rarely defined well The best definition seems to be the following: “Activities in the field that have to be done more than once in the field, or activities which remove work previously installed as part of the project regardless of source, where no change order has been issued and no change of scope has been identified by the owner.” According to Dougherty, 2013 [11]

Related Studies

Trach, Roman & Trach, Yuliia & Lendo-Siwicka, Marzena in a research 2021 [3] about using ANN to predict Rework cost causing by Communication factors was shown that 12 factors were identified and assessed that influence communication The level of rework costs in 18 construction projects was also calculated

Limitation: The study utilized a limited amount of input data, resulting in high labor costs and complexity in estimating rework costs

Moreover, the authors utilized a two-layer feedforward network architecture and tested it with three algorithms However, they recommend experimenting with multiple neural network architectures for optimal prediction accuracy

In addition, the authors understand and realize that project communication is an important factor influencing the rework cost, but still not the only one

Pudke, Milin et al with the journal in 2020 [12] about factors affecting rework cost was shown that four main group affecting to rework cost:

Table 2.1 Main group factors impact on rework cost according to Pudke, 2020

Moreover, the main factors directly impact on rework cost were lack of experience and knowledge of design and construction process, poor quality of construction technique and poor used of advanced engineering The total rework cost is approximately 15-20% of the total project cost in construction work after case study analysis

Al-Janabi, Abdel-Monem et al in a research in 2020 [14] about the factor causing and impact on rework cost in Egypt projects was identified the largest number of rework causes, as well as it provides adequate knowledge about the common causes of rework and its impact in the Egyptian construction industry

Table 2.2 Ranking the causes and impact of rework factors related to the client according to Al-Janabi, 2020

Specifications change by the client 44.41% 1

Inadequate or weak feasibility study 44.28% 2

Change of plan or scope 43.79% 3

Weak planning of the project as (construction planning) 41.58% 4 Weak communication/coordination with endusers 40.30% 5 Lack of sufficient knowledge and experience for the design and construction process 37.21% 6

Lack of funding allocated for consultation 28.37% 9

Lack of client involvement in the project 27.47% 10

Poor communication with the contractor / design consultant 24.83% 11

Table 2.3 Ranking the causes and impact of rework factors related to the contractor/subcontractor according to Al-Janabi, 2020

Lack of funding and cash flows 41.54% 1

Inefficient selection of the subcontractor 41.21% 2

Insufficient skill and defective workmanship of subcontractor 36.87% 3

Non-compliance with specifications and standards 33.51% 5

Poor coordination and communication with consultants, such as designers 32% 6

Insufficient managerial and supervisory skills 31.02% 7 Poor coordination/communication between contractor and subcontractor 28.94% 8

Restrictions for some activities and tasks 28.16% 9

Lack of safety considerations in the site 27.76% 10

Failure to protect of the constructed works 24.60% 11

Table 2.4 Ranking the causes and impact of rework factors related to the contract according to Al-Janabi, 2020

Unclear/poor items of contract documentation 33.13% 1 Insufficient time required to prepare contract documentation 31.16% 2

Inexperienced staff to prepare contract documentation 29.56% 3

Errors or omissions in the contract documentation 26.03% 6 Lack of funding by the client to prepare contract documentation 23.77% 7

Poor implementation of the contract 22.68% 8

Table 2.5 Ranking the causes and impact of rework factors related to the design according to Al-Janabi, 2020

Design change due to its contradiction with the utilities 47.83% 1 Design change at the client order or enduser 40.66% 2

Incomplete design at the time tender 38.07% 3

Errors and omission during the design process 35.69% 4 Design change because of the difficulty of implementation 35.51% 5

Poor coordination and communication between the design team 32.50% 6

Unclear of the client’s requirements 31.52% 7

Insufficient experience of the design consultants 30.52% 9 Misunderstanding of client/end-user requirements 29.61% 10 Non-compliance with specifications and standards 26.65% 11

Pay the low fee for design preparation 23.62% 12

Lack of using modern technology and software 22.78% 13

Table 2.6 Ranking the causes and impact of rework factors related to the supervisory according to Al-Janabi, 2020

Improper planning for the project activities 38.36% 1 Inadequate supervision by design consultant 33.62% 2

Lack of motivational and leadership skills 30.41% 4

Misunderstanding of schemas and designs 26.18% 6

Table 2.7 Ranking the causes and impact of rework factors related to the construction according to Al-Janabi, 2020

Changes initiated by client/end-user after work has been undertaken 41.01% 2

Inadequate pre-construction planning of the project 36% 3 Changes due to inappropriate/difficult the methods of construction in site 34.39% 4

Repair of damage caused by the subcontractor 32.90% 5 Changes made at the request by a design consultant for quality improvement 32.04% 6

Changes made at the request by the contractor to improve quality 31.26% 7

Change due to omissions or nonconformance to the original design 29.44% 8

Construction error due to design misunderstanding 24.38% 9 Lack of use proper and modern construction technology 21.68% 10

Table 2.8 Ranking the causes and impact of rework factors related to the site according to Al-Janabi, 2020

Poor site conditions (e.g., soil problems, water, electricity) 37.27% 1

Lack of client funding for the site investigation 36.46% 2 Poor management and resource planning in the site 30.89% 3

Lack of safety and job security 27.43% 4

Bad site practices and site status that is overlooked by the contractor 26.12% 5

Lack of support for site management by the contractor 22.92% 6 Delay providing the site requirements for the contractor by a client, such as water, electricity 22.10% 7

Table 2.9 Ranking the causes and impact of rework factors related to the labours according to Al-Janabi, 2020

Shortage of skilled labours required to complete work tasks 37.88% 1

Reallocation /turnover of the staff to another site 35.45% 3

Insufficient training to develop a skill 34.12% 4

Lack of incentives and rewards for staff 30.50% 6

Failure to comply with safety instructions 28.71% 7

Lack of supervision and planning for work tasks 27.88% 8

Table 2.10 Ranking the causes and impact of rework factors related to the materials and equipment according to Al-Janabi, 2020

Material and Equipment Factors % Impact Ranking

Lack of materials in a site when needed 37.60% 1

Inappropriate delivery timing of the materials and equipment by the supplier 30.53% 2

Poor supervision by the contractor for the acceptance’s materials/equipment 25.37% 6

Table 2.11 Ranking the causes and impact of rework factors related to the external according to Al-Janabi, 2020

Country’s economic situation (currency exchange, inflation) 69.68% 1

Change of laws and government decisions that initiated many changes, such as (design change) 37.40% 2

Impact of social and cultural factors 24.97% 3

Natural conditions, such as rain, heat, and cold 21.54% 4

Table 2.12 Ranking of each rework group according to Al-Janabi, 2020

Material and Equipment related factors 9.12% 7

According to Vu, Pham et al with research in 2020 [15] about “Factors

Influencing Cost Overruns in Construction Projects of International Contractors in Vietnam” was demonstrated that 31 factors divided into five main groups (Owner, foreign contractor, subcontractor, state management, project) that often lead to construction cost overruns for foreign contractors in Vietnam

Table 2.13 Ranking of contractor factors on cost overruns in construction projects according to Vu, 2020

Loose management and poor supervision of quantities 1

Weak management of contracts with owners concerning increasing quantities and remedying of the defect and unfavorable contractual terms for contractors

Limited experience of construction organization 4

Improper communications between contractors and subcontractors 5

Loose management and poor supervision of quality and schedule 7

Delay in the decision-making process of contractors 8

Loose management of site material and machinery warehouses 9

Table 2.14 Ranking of Owner and project manager factors on cost overruns in construction projects according to Vu, 2020

Owner and project manager related factors Ranking

Extension of the project implementation schedule 1

Contract suspension or arbitrary contract cancellation 2

Owners’ deliberate presentation of incomplete quantities in bidding documents 6

Table 2.15 Ranking of Design factors on cost overruns in construction projects according to Vu, 2020

Poor quality of design documents 1

Delays in dealing with design changes 2

Adoption of different construction codes and standards from those ever applicable to foreign contractors’ projects in their countries 3

Table 2.16 Ranking of engineer factors on cost overruns in construction projects according to Vu, 2020

Untimely instructions and decisions on acceptance and approval of drawings and quantities 1

Poor qualifications of site engineers 2

Table 2.17 Ranking of Subcontractor and supplier factors on cost overruns in construction projects according to Vu, 2020

Subcontractor and supplier related factors Ranking

Supply of poor-quality materials and supply delays 2

Subcontractors’ insufficient finance for construction 3

Table 2.18 Ranking of other factors on cost overruns in construction projects according to Vu, 2020

Abrupt increases in site worker demand 1

Difficulties or delays in administrative authorities’ approval and licensing process 2

Legal document changes with impacts on projects 3

Inflation or increased interest rates 5

Unclear and incomplete geological survey documents 6

Bureaucracy, corruption, and bribery of stakeholders 7

Differences in culture and working environments and language barriers 8

In 2021 on a economics and business administration journal, Tran Pham Khanh, 2021 [6] researched “Factors affecting cost overrun of infrastructure projects in

Ho Chi Minh City” showed the analysis of factors affecting cost overruns in technical infrastructure construction projects in Ho Chi Minh City The results of regression analysis show that the factors affecting the project's ability to exceed estimates are arranged in descending order as follows: Investor; contractors; consultant; economic; natural and social environment; policy

Limitation: This research just surveyed in within Ho Chi Minh City and small sample size In addtion, research just focused on general project

Table 2.19 Ranking of regulation factors on cost overruns in infrastructure projects according to Toan, 2021

Tax regulation is not stable 1

Mechanism of Bidding Law and Construction Law no clear 2

Salary regulation is not stable 3

Table 2.20 Ranking of economic factors on cost overruns in infrastructure projects according to Toan, 2021

Loan interest rates exceeded the plan 3

The price of construction materials exceeded the plan 5

Table 2.21 Ranking of nature and society factors on cost overruns in infrastructure projectsaccording to Toan, 2021

Nature and society related factors Ranking

Weather/geological/hydrological conditions at the construction site are more complicated than those in the survey 1

Other agencies do not coordinate synchronously when recovering and clearing land 2

People's protests are due to inadequate compensation and site clearance 3

Table 2.22 Ranking of consultant factors on cost overruns in infrastructure projects according to Toan, 2021

The total investment estimate is not accurate 2

The Consultant's capacity is poor 3

Table 2.23 Ranking of investor factors on cost overruns in infrastructure projects according to Toan, 2021

Poor ability to delegate authority to subordinates 2

Poor transparency and explained responsibility 7

Project appraisal is not objective 8

Poor coordination ability with participating parties 10

Table 2.24 Ranking of contractor factors on cost overruns in infrastructure projects according to Toan, 2021

Outdated and inappropriate construction methods 1

Machinery and equipment are not enough to meet construction requirements 3

Poor project management and supervision 4

The main contractor's human resources are not sufficient during construction 5

Poor coordination capacity between contractors and investors 6

On construction magazine August 2021, Truong and Cao [8] published an article about “Factors affecting "re-work" in the design stage of construction projects” with the result was proved that 50 factors were identified lead to “rework” on design stage The study also pointed out 5 main groups for factors: Investor, design team, design company, contractor, external After analysing the study found 39 causes that have a great influence on "rework", among which the most prominent are the causes of initial ideas, requirements and changes of the investor, customer, causes of process and design criteria, causes of the capacity of the parties, causes of coordination and information exchange between the parties and causes from objective factors such as changes in circulars or complex construction geology

Table 2.25 Ranking of investor factors according to Truong, 2021

The investor requested to change the design 2

The investor's requirements are not clear 3 The investor makes decisions at the wrong time or late 4 Information exchange with design consultants is unclear and ineffective 5 The investor or the investor's consultant lacks experience and understanding of design processes and standards 6

The investor made unreasonable technical requirements 7 The investor requests to add a certain detail or item 8 The investor requested an unfeasible design schedule 9 The investor requested completion earlier than the original schedule 10

Table 2.26 Ranking of Design group factors according to Truong, 2021

Design group related factors Ranking

The design team lacks design capacity and experience 1 The design does not match the requirements of the investor and customers using the project 2

Wrong design with technical requirements 3

Lack of coordination between parts of the design team 4 Communication between design departments is unclear and inconsistent 5 Not clearly understanding the design manager's requirements 6 Drawings are sketchy, difficult to understand, and do not meet technical standards 7

Not understanding and properly applying current design standards 8 The design solution is not consistent with current popular construction methods 9

Not checking drawings before submitting documents 10 Facility design is unreasonable or does not comply with standards 11 Not proficient in using design software applications 12 Not clearly understanding the Company's design process and design criteria 13

Not clearly understanding the types and specifications of materials and equipment 15

Drawings are not systematically managed and not arranged in order 16

Carry out multiple projects at the same time 17

Failure to complete work on schedule 18

The design is missing or missing one or several details or components 19

Table 2.27 Ranking of Design consultant factors according to Truong, 2021

Design consultant related factors Ranking

Not clearly understanding the requirements of investors and customers 1

Design criteria are not clear 2

No or inappropriate design process applied 3

Information exchange with design staff is not clear 4

Lack of human resources to complete a workload within the required time 6

Ineffective management and coordination of design subcontractors 7

Not using current design standards or not using current construction technology 8

There is no human resource training strategy or improper and ineffective training 10

Allocate or transfer personnel to other projects 11

Not planning to divide specific tasks 12

Table 2.28 Ranking of Contractor factors according to Truong, 2021

The subcontractor's professional capacity is poor 1

Lack of coordination between subcontractors and main contractors 2

Communication between subcontractors and main contractors is not clear and not timely 3

Subcontractors continuously change design plans 4

Subcontractors undertake multiple projects at the same time 5

Table 2.29 Ranking of External factors according to Truong, 2021

Changes in planning after having a complete design 1

Geological survey work is sketchy and inaccurate 2

Changes in circulars and standards during the design process 4

Table 2.30 Ranking of TOP 10 impact factors according to Truong, 2021

The investor requested to change the design 2

The design team lacks design capacity and experience 3

The design does not match the requirements of the investor and customers using the project 4

Not clearly understanding the requirements of investors and customers 5

Changes in planning after having a complete design 6

Design criteria are not clear 7

Geological survey work is sketchy and inaccurate 8

No or inappropriate design process applied 9

Wrong design with technical requirements 10

Comment: Currently, few studies have been focused about rework cost on

Vietnam construction sector In addition, there have been few researches determined factors causing rework cost for all of aspects of construction fields In order to collect information, this research was referred from studies related to overun cost on infrastrucure, Bridge/ road project for improving research quality Moreover, most of rework factor researches have been from overeas with a difference about geology, economic condition and construction field characteristic which could not be applied all of factors on this thesis So it is necessary to statistic and filter main factors which affect to rework cost in Vietnam construction field so that it is suitable with research purpose

On the artificial intelligence era, many current reasearches have applied AI method to predict and estimate cost and impact percentage on construction problem Support vector machine, Deep learning and Artificial neural network would be used for supporting human career Among those methods, ANN would be considered as the most suitable method applied on this study in order to predict and evaluate impact of rework cost on Vietnam construction projects through actual data from completed projects.

Chapter 2 summary

Chapter 2 outlined the definition of "rework cost" and the factors that cause it in construction projects It also introduced the concept of artificial neural networks (ANN), explaining their relevance and application in this research for data analysis The chapter referenced several international and Vietnamese articles, providing context for the study However, the current research on "rework cost" in Vietnam's construction projects remains incomplete, particularly in terms of exploring all aspects of construction and leveraging AI to address the issue In this research, the focus is on analyzing data to identify factors that influence rework costs The analysis will be conducted using ANN, with results presented in subsequent sections This approach aims to provide a comprehensive understanding of the factors contributing to rework costs and explore AI- based solutions to mitigate these costs in Vietnam's construction industry.

RESEARCH METHODOLOGIES

Research Procedures

Factors affecting Rework Costs in Construction Projects

Identify factors cause Rework items in Construction field Experts opinion, experience reference Science Articles, Engineering

Trial survey and expert opinion reference and revision

Data Collection and Analysis (Ranking, test)

Use results for 2nd Questionnaire

After determining research procedures, starting to refer from previous research articles, rework factor information have been collected from construction experts who have been teaching on universities and working on material suppliers and contractor and experienced member from work on construction environment by discussing and expressing to them with thesis outline and procedures Basic questionnaire design and processing trial survey Revising questionnaire form following to experts comments and feedback from trial survey Official and widespread survey (By hard copy and e-form) Then carrying out data collection and classification Data Analysis for ranking and testing collected data On next step, giving comments, discussions and solution for issue Results from SPSS part would be used for designning 2nd Questionnaire form Together with collected data, those data would be applied for input and output data for ANN Modelling Running software with ANN model and other models, output results from software would be evaluated and discussed Finally, author give comment and discussion for graduation project.

Questionnaire Survey Purpose

The main purpose of survey is that support researcher collect data from respondent (Investor, Consultant, Contractor, Supplier on construction field) and use collected data in order to analysis, evaluate results Then those results will be used for discussion and making solution for research topic With advantages such as low cost, short time for collecting, large size of respondent, easy for design, which have been used for researchers, students gain data to serve for final report.

Questionnaire Design

In the beginning, idea for questionnaire is very important When idea have been created, previous researches and experienced experts shall be referred for proper content of questionnaire table

According to previous research such as Al-Janabi, 2020 [14], Forcada, 2017 [16],

Vu, 2020 [15], Le-Hoai Long, 2008 [5], Toan, 2021 [6], Truong, 2021 [8] research issues would be combined and developed a preliminary survey questionnaire

On questionnaire design period, totally 7 groups with 46 factors those would be used for surveying Then this questionnaire table would be sent to subjects such as Project

Manager, Site Engineer, Architectural Engineer, Structural Engineer, Experts on Construction Field

After revising, questionnaire form would be sent to people from Investor, Consultant, Contractor and Supplier by hardcopy and e-form.

Questionnaire Content

This part will be help respondents understand about the purpose, information and privacy commitment

- Part I: General Information- which related to respondent career and experience such as position, experience year, project type

- Part II: Group of causes that affect rework costs of construction projects

- Part III: Information for percentage of cost increase due to rework in comparison with construction cost estimation

Measurement scales: Likert 5 measurement scales would be applied to evaluate occurrence level and influence level of each factors cause rework

Answer selection: Each of factor would be ranked with 2 required levels

In construction projects, the most important factors to evaluate the level of effectiveness include Time - Cost - Quality Cost is the most powerful factor and causes many impacts on time and quality Currently, many construction projects are completed with budgets exceeding the investment budget, which leads to a lot of controversy and loss of trust from users The number of causes of cost increases in projects and among them the costs of "rework" construction items are very concerning This survey aims to identify and analyze factors affecting "rework" costs of Vietnamese projects

According to previous research articles such as Al-Janabi, 2020, Forcada, 2017,

Vu, 2020, Le-Hoai, Long, 2008, Toan, 2021, Truong, 2021, “rework” items have been divided into some group such as Investor, Consultant, Contractors, which included many factors appeared frequently on construction project all around the word

Totally 7 main groups which were identified as aspects impact on rework cost With 46 factors could be huge impact on rework cost Especially some featured factors of Vietnam construction field According to contribution and advice from instructor, questionnaire form was revised with 3 additional factors During working experience on current project, 13 factors have been identified as the most frequent factors on construction duration

Figure 3.2 Groups impact Rework cost

In order to serve for data collection into SPSS software, factors impact on rework cost would be encode on below table:

CODE Factors impact on rework cost Reference

CDT1 Poor communication and connection with design parts/contractors [17]

CDT2 Decision making at the wrong time or late [8]

CDT3 Plans, designs, and adds scope of work changes Personal

Factors Environment and External Factors

CDT4 Lack of experience and understanding of the design and construction [17]

CDT5 Financial and capital issues [5]

TK1 Incomplete design documents at the bidding stage Personal

TK3 Too complicated design details, leading to difficult construction [17]

TK4 Unclear and inconsistent communication and connection between members between design departments [12]

TK5 Design details and structure missing in items Personal

TK6 Inconsistent and incompliant design with standards and regulations [8]

TK7 Vague and unsystematic design documents [8]

C Factors related to Supervision Consultant

GS2 Incomprehensive clearly project design documents Personal

Experience GS3 Poor organization and supervision of project work [14]

GS4 Inexperienced understanding regulations and information related to construction in TCVN

Personal Experience GS5 Inspection and control of design errors delay [5]

NT1 Insufficient financial capacity to meet project requirements Personal

NT2 Incompliant with design technical requirements Personal

NT3 Poor communication and connection with project managers and supervisor [14]

NT4 Poor quality construction materials choosing [17] NT5 Omissions, skipping steps in construction methods [17]

NT6 Poor capacity subcontractors selection [14]

NT7 Inefficient protection for completed items [17]

NT8 Lack of connection and unity with the Investor and consultants

E Factors related to the environment and external factors

MT1 Bad and extreme weather [15]

MT2 Natural disasters, war, epidemics [18]

MT3 Protests and disputes over compensation and clearance issues [18] MT4 Complicated geological situation than the survey [6]

F Factors related to labour and equipment

LD1 Skilled workers shortage to complete the work [14] LD2 Inefficient safety guideline workers during construction [14] LD3 Ineffective training and developing skills for workers [14] LD4 Outdated machinery and construction technology [14]

LD5 Broken and poor quality construction equipment Personal

KH1 Changes in laws and government decisions led to many changes in construction [18]

KH2 Outdated design due to long construction period Personal

KH3 Obstacles in permit procedures related to other technical infrastructure projects [18]

KH4 Errors or omissions in terms and scope of work in the contract

KH5 Technical measures and marketing technology that are too new [18]

KH6 Ambiguous payment conditions between parties Personal

KH8 Quality construction materials shortage in some localities in

KH9 Law ovelaps law in the Vietnamese legal regime Expert

KH10 Low disbursement of capital sources Expert

Through questionnaire form, the final part which help collect percentage of rework cost in comparison with total cost in completed project from respondents This data information would be used to analyze ANN model at next part of this report.

Data Collection

Sample Size would be identified to know quantity of survey needed to collect According to Hair, Black, Babin, và Anderson (2010) [19], sample size need to be 5 times of observed variable On this research, observed variables are 46, so the sample number can be 230 or more

However, period for surveying is not enough and from previous researches, sample size from 120-130 would seem to be adequated for research

Data collection would be carried out by sending hardcopy questionnaire form or e- form to respondents from Investor, Consultant, Contractor, Supplier E-form would be sent e-form via Google form and email link After collecting and checking data, all of unsatisfied answer would be removed

Nowsaday, ten of thousand of social media platforms with information transmission, some platform such as Messenger (Meta), Zalo, Skype, Gmail would be used to send e-form link to respondents at the same time with hardcopy that have been handed out to colleague and classmate.

Data Analysis Tools

SPSS Statistics 25 would be applied for:

 Statistical Questionnaire Analysis (Table, Chart, Graph)

 Cronbach’s Alpha Coefficient Analysis (Reliability of the measurement scale checking)

 One-way ANOVA or Kruskal - Wallis: Using for evaluating whether or not a difference in the mean between the groups responding to the questions

 Exploratory Factor Analysis (EFA): Using for grouping and combining variables

Microsoft Office Excel 2016 would be used for describe statistics and result analysis from SPSS

Rapidminer Studio 95 would be used for creating the predictive models for estimation and evaluate model (R, MAPE, MAE, RMSE)

3.6.1 Cronbach’s Alpha Coefficient (Reliability Analysis):

Testing the reliability of Cronbach's Alpha scale is used to eliminate useless variables This step is very important to the reliability of the questions as well as the analysis results Cronbach's Alpha coefficient is a statistical test of how closely the items in the measurement scale correlate with each other

According to Hoang and Chu (2008) [20] Cronbach’s Alpha Formula is:

N: Number of questions ρ : the average correlation coefficient between the observed variables

Hoang and Chu (2008) also mention that a good evaluation for a measurement scale is α ≥ 0.8

No Cronbach’s Alpha Evaluation Level

3.6.2 Factors ranking due to mean:

Ranking method depend on mean would be used in thesis for classifying the important factors impact on rework cost on construction project through questionnaire survey data

One-sample test is used to compare the sample mean with a specific value In this study, the one-sample would be used to check whether the survey subjects are higher than the neutral level or not

In other to analyse the difference about mean of general groups, One-way ANOVA is suitable for sample testing Due to period and condition, this data analysis are not certainly exact so One-way ANOVA and Robust test could be applied together for comparing results (ANOVA for Sig Levene’s test ≥ 0.05 or Robust for Sig Levene’s test

< 0.05) If both of sample testing are sample, the results will be reliable with assumptions:

H0: No differences in mean levels occurred between groups

H1: Differences in mean levels occurred between groups

Exploratory factor analysis EFA is used to determine convergent validity, discriminant validity and shorten of estimated parameters for groups of variables according to Le Dinh Hai (2018) [21] Moreover, EFA is also used for evaluating preliminarily measurement scale according to Nguyen Dinh Tho (2014) [22]

According to Tho 2014, Bartlett’s test is used to check whether the correlation matrix is identity matrix, a matrix with components (correlation coefficients between variables) equal to zero and diagonal (correlation coefficient with itself) equal to 1 If Bartlett’s test with p < 5%, Hypothesis Ho would be rejected, this means that the variables are related to each other

KMO (Kaiser - Meyer – Olkin measure of sampling adequacy) is an index used to compare the size of the correlation coefficient between two variables with the size of their partial correlation coefficients In other to use EFA, KMO have to be ≥ 0.5

Eigenvalue criteria: The number of factors is determined at the Eigenvalue factor with ≥ 1 would seem to be extracted factors have summary value that represent good information

Total Variance Extracted > 50% would seem to be accepted value

According to Hair et al.(2010), Multivariate Data Analysis, 7 th Edition about

 Factor Loading (FL) in the range of 0.3 to 0.4 would be meet the minimum level

 Factor Loading (FL) in the range of 0.5 would be considered partically significant

 Factor Loading (FL) in range of 0.7 would be considered well-defined structure

In case of sample size from 100-150 The Factor Loading would be in range of 0.45-

Network architecture refers to the arrangement and connection of neurons in a network, and can be represented by various ANN architectures such as multilayer perceptron (MLP), generalized feed-forward neural network (GFNN), support vector machine (SVM), generalized regression neural network (GRNN), radial basis function neural network (RBFNN), neuro-fuzzy and others (R.Trach, Y.Trach, Siwicka 2021) [3]

Artificial Neural Networks (ANNs) are powerful tools that predict efficacy by imitating human neural systems They learn from past data and consist of three layers: input, hidden, and output The input layer uses variables, hidden layers calculate them, and the output layer provides the predicted factory construction cost (Chou et al 2022) [23]

Figure 3.3 A typical ANN architecture according to Srivastav, Roshan & Sudheer, K &

Compared to traditional networks, Artificial Neural Networks (ANNs) provide a number of benefits, including enhanced generalization capacity, improved learning efficiency, improved pattern recognition accuracy, and the capacity to manage noise and imperfect input ANNs can also process a much larger amount of data than classical networks Artificial neural networks have been used extensively in building recently ANNs have been particularly effective in risk assessment, schedule optimization, cost forecasting, and dispute settlement ANNs are used to address issues that are challenging for conventional statistical and mathematical approaches to solve

This study utilized a learning algorithm to determine connection weights between neurons, training a network using error backpropagation, and dividing data into training, validation, and testing sets

The multilayer perceptron is a feedforward ANN model that maps input datasets to output datasets, achieving specific prediction tasks like object identification It consists of input layers, hidden layers, and output layers, with neurons connected to each other, allowing for efficient data analysis Inputs can be image or document feature values, while outputs achieve specific tasks (Xihe 2022) [24]

Backpropagation method to help calculate the gradient layer by layer backward from the last one In addition, backpropagation divides error among connections, advancing ANN algorithmic perspective, reduces repeat calculation, and significantly enhances gradient searching efficiency The BP method is mainly based on the derivative chain rule [24]

According to R.Trach, Y.Trach, Siwicka 2021 [3], at the first stage, each neuron in the input layer receives an input signal and transmits this signal to each neuron in the hidden layer

Then each hidden neuron calculates its activation and transmits its signal to each output neuron For a neural network, the total input signal is determined by the equation

Where: xi= input variables wi= weights

Each output neuron calculates its activation to form a network response to a given input signal The activation function determines if a signal's summed result can produce an output signal, often associated with hidden layers neurons, and is frequently utilized for activation Many types of activation function according to Ge 2022 [24]:

 Step function is challenging to use in perceptrons and ANN development, leading to the discovery of numerous variant activation functions with improved performance

 Sigmoid function is a commonly used "S" type function, which can map variables to the interval (0,1) Sigmoid functions are widely utilized activation functions in multi-layered feed forward neural networks

 The rectified linear unit function (ReLU) aligns with physiological models, avoiding exploding gradients and accelerating convergence It simplifies calculations and always equals 1 if the input is positive: R(x) = max(0, x) The validation stage, which begins after network training, is crucial for the network's successful use Its objective is to ensure the network can generalize data within the training boundaries, and if these characteristics are adequate, the model is considered valid

Figure 3.4 Step function, Sigmoid function, its derivative and ReLU function

The final stage of testing is verifying how the network works with data that it has never seen before during its development

Prediction accuracy would be measured by RMSE (Root Mean Squared Error)

Model Evaluation

Model through Rapidminer would be evaluated according to performance evaluation criteria:

 Mean absolute percentage error (MAPE):

 Root mean square error (RMSE):

𝑛 𝑖=1 where ya represents an actual value; yp denotes predicted value; and n is number of samples.

Rapidminer Studio 9.5

Rapidminer is one of the most analysis system and estimation which have been developed by Rapidminer company and is written by programmers in the Java programming language Moreover, it has helped provide an integrated environment for deep learning, text mining, machine learning as well as predictive analytics

Rapidminer Studio: this module is used in designing workflows, prototyping or validation processes, etc

Application: Rapidminer have been applied in many application such as business application, commercial application, education, research and development application with fuctions:

 Load and transform data system (Extract, Transform, Load (ETL))

 Systematic data processing and data visualization

 Build forecasting, analytical and statistical models

 Evaluate the data then implementing the data

 Model analysis result export rapidly

 Easy for selecting and comparing charts

 Automatically calculate model evaluation result

Chapter 3 Summary

Throughout chapter 3 is a description and interpretation of the research methods used to collect and analyze data in this master's thesis topic The first is the idea generation stage based on previous research articles, from which the survey questionnaire is developed and designed Conduct a test survey and consult instructor about the content Then determining the survey sample size, refine the questionnaire, and send the survey widely Total of 7 groups and 46 factors to survey, the results collected after sending 144 the survey forms is 125 adequate questionnaires The next step will use the results to apply to SPSS to run analysis steps On the other hand, Chapter 3 also introduce about Rapidminer software which would be used for ANN Modelling and model evaluation.

DATA ANALYSIS

Descriptive statistics

Total 148 questionnaire form were sent to respondents who were working on Construction Company, Students from Construction Manangement Master Course at HCMUT, Civil Engineer who gradutated from International University from 2019, Korean engineer experts working on Thu Thiem Zeit River Project Furthermore, questionnaire form was distributed to engineer at different regions in Vietnam, so the result would be representative for Vietnam projects

Respondents who were engineer and staff from Thu Thiem Zeit River Project parties: Investor- Vietnam GS Enterprises, Design Consultant- MAP, Supervision Consultant- ICIC, Contractor- IVY Build, UDIC, Sub-Contractors and Suppliers

Date for surveying lasted from 06th March 2024 – 31st March 2024

Data Collection Results

Table 4.1 Questionnaire response statistic table

According to table 4.1, questionnaire collection results are 129 After filtering collected data, with total of 125 adequated questionnaire responses and 4 inadequated questionnaire responses such as only selected one level (Occurrence or influence), not finished 100% all of questions All of received hardcopy questionnaire were input on google form for combining on one source About inadequated questionnaire, it were lack of fullfiling information and not enough answers were recorded With all of adequated responses which would be used for general information statistic and applied SPSS for testing and analyzing

According to percentage information, the number of respondents with 3-5 years experience occupied 15.2% and 5-10 years are 28% Futhermore 10 years experience respondents are recorded with 47.2% As the result, Engineer with more than 5 years experience would contribute a overall situation and aspects about Vietnam construction field Because the wide range of respondents on this research was Class I construction project, many experience engineers have been working and contributing their capability on project With more than half of respondents had experience larger than 5 years, the collected data could be considered as valuable data for analysis

< 3 years 3 to 5 years 5 to 10 years > 10 years

Company Role Frequency Percentage (%) Cumulative

Overall, the data indicates that invester with 43 responses occupied 34% and Design/ Supervision Consultant are 12% Moreover, engineer replied questionnaire from Contractor with 62 responses represented 50 % On the other hand, State departments and Other (Supplier, HSE) respondents only account for 4% Depend on those results, the collected data are reliable enough for next survey analysis Contractor could be considered as party play most important role on each construction project, so with portion is 50% the collected data were seemed to be the representatives of the whole picture of Vietnam construction sector

In general, it can be seen that Engineer/ Construction experts accounted for the highest rate of respondents with 83 responses (66.8%) Director and Manager/ Deputy Manager of Project are 20/125 respondents occupied 2.4% and 13.6% respectively, while 16/125 respondents from Head/Deputy Head Department Thus, the working position data is suitable for report Generally, nearly 98% respondents on this research were engineers

Director Project Manager/ Deputy Manager

Head/ Deputy Head Department Engineer/ Construction expertOther and construction experts so the collected data could be ensured for major perspective in construction business

4.2.5 Number of participated project result:

Table 4.5 Number of participated project table

In particular, the number of respondents with experience in 2-4 projects and > 7 projects occupied for highest portion with 28% and 39%, respectively On the other hand, respondents with < 2 projects experience only accounted for 6% and 5-7 projects experience were 27% Totally, more than 50% (39% respondents with experience > 7 participated projects) of respondents joined more than 5 projeccts, which showed that respondents have had wide vision for many problem which have been happend during their work experience According this result, respondents with project experience are proper for this data statistic

Table 4.6 Participated Project type table

Project Type Frequency Percentage (%) Cumulative

It is noticeable that Civil Project and Industrial Project accounted for the largest proportion 72,8% and 14,4% respectively Whereas, Bridge/Road Project occupied 4,8% and Infrastructure comprised 8% with total percentage On this research, the author do not focus on particular project, but Civil project such as building, resident houses, school and hospital are received more attention With more than 70% of respondents have been participated on Civil project, the result data would be considered more efficient and comprehensive for rework factors Rely on this result, the collected data was enough project types in construction field, so the result would be better

Civil Project Bridge/ Road/ Irrigation ProjectIndustrial Project Infrastructure Project

Table 4.7 Participated Project Scale table

Project Scale Frequency Percentage (%) Cumulative

The chart showed that the respondents project scale in the range from 100-500 billion VND with 54/125 answers, occupied with the highest rate 43,2% Next, scale of project from >500 billion VND was recorded with 39/125 respondents accounting for 31,2%, and project with scale < 50 billion VND has 7/125 responses showing for 5,6%

In contrast, respondents answered for project scale 50-100 billion VND only occupied 20% in total Generally, more than 50% respondents replied with participated projects >

100 billion VND, so it mean that they experienced on large and huge scale of projects, important project for nation and wide scope of work Thus, the results of project scale would be seem appropriate for statistic

< 50 billion VND 50-100 billion VND100-500 billion VND > 500 billion VND

4.2.8 Affection of rework on construction project:

Rework affection Frequency Percentage (%) Cumulative

It is clearly to see that most of respondents they experienced and concluded that their participated projects have been affected by rework with 117/125 answers, accouting for 94% of total Generally, almost projects around the world would occur “rework”, Vietnam construction field also have been recorded most of projects occured “rework” That is the reason why the most portion (94%) which respondents agreed that their participated projects happen rework items during construction stage and it will impact on the total amount of projects So, this questionnaire survey work is useful for applying on next analysis steps and become a basis for connecting to the ANN Modelling chapter

4.2.9 Rework cost percentage on construction project:

Table 4.9 Rework cost percentage table

Rework cost % Frequency Percentage (%) Cumulative

As on the table and chart indicated clearly, 68 respondents occupied 54,4% answers that participated projects were affected 70% only accounting for 1,6% and 0,8% respectively With the huge portion of rework cost percentage < 50% is nearly 98%, it mean that most of construction project with the rework cost will be range from 20% to 1% Especially, a project have been considered as successful project when this project are recorded with rework cost less than 5% According to this part, range of large number of projects would be have the rework cost range from 0,5, it mean that it is suitable for using factor analysis on this research

 Barlett’s test has Sig= 0.00 < 0.5, it mean that all of observed variables are correlated with each other in factor, so factor analysis is proper

Table 4.31 Total Variance Explained Table

Initial Eigenvalues Extraction Sums of

Rotation Sums of Squared Loadings Total

% of Varian ce Cumula tive % Total

% of Varianc e Cumula tive % Total

Extraction Method: Principal Component Analysis

Comment: Following to Total Variance Explained table above:

 Eigenvalue is 1,564> 1, which mean that from 43 observed variables would be summarized to 7 extracted factors

 Total Variance Explained: Such 7 factors would be explained to 71,496%

Table 4.32 Rotated Component Matrix Table

Extraction Method: Principal Component Analysis

Rotation Method: Varimax with Kaiser Normalization a Rotation converged in 6 iterations

Comment: According to the rotated component matrix, 43 factors were divided into 07 component groups, all observed variables have factor loading > 0.5 and variables with low factor loading were not appeared Overally, observed variables have been represented for actual factors causing rework cost on each construction project in Vietnam Through the rotation component analysis, each group of variables have been grouped and named following to feature and characteristic of each variables, detail results are discrible at Appendix V

4.7.2 Analyze the meaning of the main components that affect the rework costs of the project:

The first component group: Factors related to others and Vietnam specific characteristics which include 7 factors

Table 4.33 The first component group table

Code Factors Influence level mean

KH10 Low disbursement of capital sources 3,58

KH1 Changes in laws and government decisions lead to many changes in construction 3,65

KH9 Law overlaps law in the Vietnamese legal regime 3,61

KH3 Obstacles in permit procedures related to other technical infrastructure projects 3,72

KH6 Ambiguous payment conditions between parties 3,62

KH2 Outdated design due to long construction period 3,60

Factors such as “Law overlaps law in the Vietnamese legal regime”, “Low disbursement of capital sources” and “Changes in laws and government decisions lead to many changes in construction” were considered as factors occur specially on Vietnam construction field, which have been mentioned on social media so far and become challenges on national and state-run construction projects There have been some policies from government related to financial regulation and specific mechanisms of each region, as a result capital sources are slowly distributed to projects Law system in Vietnam have been updated every year which will be suitable with each stage of development and current condition matters As the consequence, Construction standard also updated according to changed regulation and some of construction work have to be rework for following law

In addition, factor such as “Outdated design due to long construction period” was related to some problem which causing project delay As a result, design applied for those projects have been gradually became outdated and unsuitable with development of society and economy Moreover, factor “Obstacles in permit procedures related to other technical infrastructure projects” is factors often occur in most of project in Vietnam, with this problem management boards need to find method and connection with related parties for completing authority document and applying for connection

Factor “Ambiguous payment conditions between parties” is always happen between contractor and investor during construction payment times Many element affected to payment progress such as Variation order, unclear scope of work, missing construction item, Non-Conformance-Report,… Furthermore, last factor “Lawsuit, strike” is considered one of the potential problem on each of project Some of popular problem such as Contractor with their manpower, Investor with local public, Main contractor with investor As a result, construction period and quality will be affected seriously

The second component group: Factors related to Design Issue which include 8 factors

Table 4.34 The second component group table

Code Factors Influence level mean

TK5 Design details and structure missing in items 3,73

TK6 Inconsistent and incompliant design with standards and regulations 3,60

TK1 Incomplete design documents at the bidding stage 3,77

TK7 Vague and unsystematic design documents 3,66

TK3 Too complicated design details, leading to difficult construction 3,69

Unclear and inconsistent communication and connection between members between design departments

Five of factors “Design details and structure missing in items”, “Inconsistent and incompliant design with standards and regulations”, “Incomplete design documents at the bidding stage”, “Vague and unsystematic design documents” and “Too complicated design details, leading to difficult construction” are often occur during design preparation Design document is the most important document for all projects, but most of project have failed for design document control With many problem happing during design stage and it will lead to construction problem on construction stage such as lack of design detail, complex design, unclear explanation, wrong specification guidance, unfollowed design standard and unsystematic document arrangement Moreover, Design consultant and Investor work on outdated design software, so the design calculation is not suitable with current standard and design detail will be showed in difficult control

Factor “Inexperienced design consultants” and “Unclear and inconsistent communication and connection between members between design departments” are two factors related to Design consultant unit They select incapacity design consultant, as a result, rework cost would be happen due to design dossiers lack of information or good material quality Furthermore, during construction duration, due to lack of communication and connection between design department, with number of serious mistake and misunderstanding opinion of each party As a consequence, construction works are damaged and need to spend cost and labour for repairing because cooperation between parties are not efficient

The third component group: Factors related to Contractor/Subcontractor which include 7 factors

Table 4.35 The third component group table

Code Factors Influence level mean

NT2 Incompliant with design technical requirements 4,04

NT3 Poor communication and connection with project managers and supervisor 3,94

NT5 Omissions, skipping steps in construction methods 3,92

NT6 Poor capacity subcontractors selection 3,97

NT1 Insufficient financial capacity to meet project requirements 4,16

NT7 Inefficient protection for completed items 3,85

NT4 Poor quality construction materials choosing 3,76

Two factors “Incompliant with design technical requirements” and “Omissions, skipping steps in construction methods” indicated that unfollowed design documents by contractor often happen on construction site and it is considered as serious impact on construction quality As a consequence, there are many item have to be reworked due to inspection from authorities department

In addition, factors such as “Poor capacity subcontractors selection” and

“Insufficient financial capacity to meet project requirements” explain that problem on selecting award bidder without careful evaluation and unclear procedure It will lead to a lot of troubles happing during construction stage and will affect on construction schedule Moreover, one of factors lead a successful project is finance Contractor need to ensure for financial platform to maintain construction activities and monthly payment for parties

If any financial problem, construction schedule will be affected and cause undesired problem such as lawsuit, strike, protest,…

“Poor quality construction materials choosing” is one of problem responsibility from contractor due to unqualified and unsuitable material technical submission Contractor also lack of consulting investor for material approval As a result, many poor and unquality material applied on construction items and lead to unsafe for users in the future On the other hand, “Poor communication and connection with project managers and supervisor” is frequent factor causing unclear opinion, disunity and argument between PMU, contractor and consultant Eventually, there have been many rework items due to misunderstand and conflict idea between parties

The fourth component group: Factors related to Investor which include 6 factors

Table 4.36 The fourth component group table

Code Factors Influence level mean

CDT3 Plans, designs, and adds scope of work changes 3,91

CDT1 Poor communication and connection with design parts/contractors 3,98

CDT2 Decision making at the wrong time or late 3,74

CDT4 Lack of experience and understanding of the design and construction 3,80

CDT5 Financial and capital issues 3,032

Both of factors “Decision making at the wrong time or late” and “Financial and capital issues” are often occur with Investor activity, especially financial and capital problem is one of the serious problem causing delay and rework in Vietnam project due to lack of plan for cash flow system, unsuitable capacity mobilization and uncontrollable debt In construction project, some works will be take long time for final decision firm Investor such as design concept, material, construction method, authorities issue, because they need to cooperate with parties for reaching good condition As a consequence, construction schedule will be delayed in comparison with approved schedule

During construction period, it is difficult to avoid change and additional work because actual construction will appear much of problem which will conflict with design concept As a result, Investor need to discuss with parties for improving, adding and changing design and construction plan for better completion condition

Both of factors “Poor communication and connection with design parts/contractors” and “Lack of experience and understanding of the design and construction” are considered as inexperience from Investor They are lack of effective communication method with parties and can not manage related parties participating project through construction stage Moreover, they lead project with insufficient information about construction work and procedure Consequently, problem with rework will be occurred frequently and affect to project faces such as cost overrun, construction schedule extension, uncompleted construction works,…

The fifth component group: Factors related to Labour and Equipment which include 5 factors

Table 4.37 The fifth component group table

Code Factors Influence level mean

LD3 Ineffective training and developing skills for workers 3,49

LD1 Skilled workers shortage to complete the work 3,55

LD4 Outdated machinery and construction technology 3,34

LD2 Inefficient safety guideline workers during construction 3,51

LD5 Broken and poor quality construction equipment 3,50

Manpower is the indispensable factor on construction field However, ten of thousand of issues related to manpower and effect to project Factors such as “Ineffective training and developing skills for workers”, “Skilled workers shortage to complete the work” and “Inefficient safety guideline workers during construction” are showed that worker skill will contribute an important part on construction success, so with high-end project, high skill workers are recruited for in charge of construction work In addition, through construction stage, number of classes which organized by contractor for improving and training worker skill for ensuring quality construction finishing Moreover, safety is the first factor need to be ensure for any project, it have been recorded that many cases that have to suspend due to safety problem That is the reason why contractor strictly follow and control 100% safety issue for construction activities and labour work

Construction equipment and machinery are also consider impact factors on construction quality such as “Outdated machinery and construction technology” and

“Broken and poor quality construction equipment” Those factor can lead to unwanted issue related to poor construction quality, safety issue and working capacity As a result, reworks are occurred frequently during construction stage with cost and quality influence

The sixth component group: Factors related to Supervisory which include 5 factors

Table 4.38 The sixth component group table

Code Factors Influence level mean

GS5 Inspection and control of design errors delay 3,67

GS4 Inexperienced understanding regulations and information related to construction in TCVN 3,64

GS2 Incomprehensive clearly project design documents 3,65

GS3 Poor organization and supervision of project work 3,69

Chapter 4 Summary

With 125 collected survey questionnaire, nearly 50% respondent are > 10 years experience and almost 95% respondents are experts/engineers in construction field Moreover, respondents with experience on civil project occupied with largest portion in total (72.8%) According to collective data and ranking on Table 4.17 and 4.18, occurence level and influence level mean were evaluated with range from [2,640÷ 3,344] and [2,784÷ 3,624] so with measure scale Likert 5, the surveyed resulted was acceptable After testing for Cronbach’s Alpha coefficient and EFA, total of 7 component groups which have been represented for all of factors used on this research As a consequence, the component groups were explained following to characteristics of each factors including others and Vietnam specific characteristics, Design Issue, Contractor/Subcontractor, Investor, Labour and Equipment, Supervisory, Environment and external impact which component groups would be used for surveying data on the next chapter: ANN Modelling.

ANN MODELLING

Modelling progress

ANN Model would be set up and analysed following to 3 main steps:

Figure 5.1 Three main steps chart for modelling progress

On the 1st step, Data about construction project (Project information, contract type, contract amount, rework cost, ) were collected from 86 projects Those data would be divided into 5-fold in which 4-fold will contain 19 datasets and 1-fold will contain 20 datasets Every running model, 4-fold would be used for training and 1-fold would be used for testing

On the 2 nd step, Datasets would be input into Rapidminer Studio 9.5 Software and it would be design for input layer, Hidden layer and Output layer

On the 3 rd stage, ANN Model evaluation with correlation coefficient (R), mean absolute percentage error (MAPE), root mean square error (RMSE), and mean absolute error (MAE) The more high R is, the more strong relationship is.

ANN Modelling parameter

In order to design a ANN Modelling, some criteria need to be determined and apply for model such as Hidden layer, hidden number, learning rate, momentum, So it is one of the most challenged when choosing ANN Parameter for running

5.2.2 Hidden Layer and neuron number:

According to Uzair & Jamil 2020 [25], The hidden layer is an intermediate layer between the input and output layers of a network, with the number of hidden layers varying depending on the problem Some cases single hidden layer is applied, some situation multiples hidden layer is considered is good choice These layers consist of small neurons that transfer data from the training layer to the output layer

Data Collection and Data RefineANN Model EstablishmentAnalysis and Evaluation

Hidden layers selection is seem to be a challenged step because in some cases ANN Model is shown with overfiiting and underfitting results related to number of hidden layers As the consequence, those occurs will can significantly impact the network's efficiency and time complexity Overfitting condition occurs when the number of hidden layers exceeds the complexity of the problem, leading to overtraining and affecting the network's time complexity This condition occurs when the network matches data too closely, causing it to lose its generalization ability over the test data Underfitting condition occurs when the number of hidden layers in a network is less than the complexity of the problem, causing the network to struggle to handle such problems, negatively impacting its efficiency and resulting in inefficient results of the network

Also on this research, the research tables about hidden layers related to previous researches is combined about comparison between accuracy and time complexity of different Neural Network with different hidden layers number Most of researchers have shown that one or two hidden layers would be consider for good accuracy and less time complexity with small data So on this research, single hidden layer would be applied for designing and calculating

The number of researchers have shown demonstrated formula for determining the number of hidden neurons in ANN such as Li at al, 1995, Tamura & Tateishi, 1997, Shibata & Ikeda, 2009, Sheela & Deepa, 2013 … According to the comparative analysis research of Vujicic, Matijevic et al 2016 [26] was proved that formula of Sheela & Deepa,

2013 with the least error during training model duration [27]:

𝑁 𝑖 2 − 8 Where: Nh: Number of hidden neurons, Ni: Number of input neurons

On this research, Input neurons were determined are 14 so following to above formula the number of hidden neurons are 4

The learning rate is a hyper parameter that regulates the model's response to the estimated error each time the model weights are updated The learning rate is a configurable hyper parameter used in neural network training, typically having a small positive value between 0.0 and 1.0

The learning rate is a crucial factor in determining the speed at which a model is adapted to a problem Smaller rates require more training epochs due to smaller weight changes, while larger rates result in rapid changes and require fewer epochs A large learning rate can lead to a suboptimal solution, while a small rate can cause stuckness

Momentum in a neural network aids in faster convergence to a good solution during training by carrying over previous weight updates, enabling the network to escape local minima and saddle points

In a neural network, though, an excessive amount of momentum can be harmful Overshooting the global minimum might lead to oscillations or divergence in the network during training if the momentum is set too high This could result the network's convergence and lead to bad performance

To avoid these issues, it is crucial to set the momentum hyperparameter carefully The optimal value depends on the dataset and network architecture Tuning the momentum hyperparameter during training is recommended by trying different values and observing their impact on the network's performance Generally, a momentum value of around 0.9 is used So on this research, momentum on ANN model would be applied value 0.9

According to Vijay Kotu, Bala Deshpande, in Data Science (Second Edition),

2019, [28] Training cycles is the amount of repetitions during a training cycle It is defaulted to 500 Because the prior weights in a neural network are so varied each time a training record is evaluated, the cycle must be repeated several times So on this research training cycles would be apply with 500 cycles

Data collection was gathered from completed building, industrial, residential projects from 2013-2023 During 3 years working on bidding department of GS E&C in the main role of Investor with more than 20 bidding packages were held The author built good relationship with engineers from reputated construction company in Vietnam Total collected data is 86 project database with respondents from engineer have worked on

Construction management board, bidding department,… Data were collected during contract agreement information and via communication via online conservation and interview According to analysis result from Chapter 4, total 7 component groups which would be became data for surveying about the rate of factor impact on each completed project The collective database of 86 projects with 15 related information as table below:

Table 5.1 Data collection for rework cost

Sign Data name Unit Types

X5 Impact level of Investor’s factors NA Discrete

X6 Impact level of Contractor/Subcontractor’s factors NA Discrete

X7 Impact level of supervisor’s factors NA Discrete

X8 Impact level of Design Issue’s factors NA Discrete X9 Impact level of environment and external factors NA Discrete X10 Impact level of labour and equipment’s factors NA Discrete

X11 Impact level of others and Vietnam specific characteristic’s factors NA Discrete

After collecting datadase, it would be handled and classified such as:

 Project location: collected data include 13 cities and provinces in Vietnam: (1) Ho Chi Minh City, (2) Hanoi, (3) Ba Ria- Vung Tau, (4) Hai Phong, (5) Binh Duong, (6) Hung Yen, (7) Dong Nai, (8) Khanh Hoa, (9) Quang Ngai, (10) Bac Binh, (11) Quang Nam, (12) Nghe An, (13) Hau Giang, (14) Da Nang, (15) Vinh Phuc, (16) Ha Nam

 Construction Type: Construction purpose of each project: (1) Building, (2) Industial Project, (3) Residential project

 Construction scale: Total area of each project with range [min-max] [3500÷ 600000]

 Complex Rate: the complexity for Architecture, structure, scale, story number, of each project with level [min-max] = [1÷ 5]

 Contract type: (1) Lumpsum Contract, (2) Unit price contract

 Impact level of Investor’s factors: the influence level of factors related to Investor which causing rework cost on each project with level [min-max] [1÷ 5]

 Impact level of Contractor/Subcontractor’s factors: the influence level of factors related to Contractor/Subcontractor which causing rework cost on each project with level [min-max] = [1÷ 5]

 Impact level of supervisor’s factors: the influence level of factors related to supervisory which causing rework cost on each project with level [min- max] = [1÷ 5]

 Impact level of Design Issue’s factors: the influence level of factors related to Design Issue which causing rework cost on each project with level [min- max] = [1÷ 5]

 Impact level of environment and external factors: the influence level of factors related to environment and external factors which causing rework cost on each project with level [min-max] = [1÷ 5]

 Impact level of labour and equipment’s factors: the influence level of factors related to labour and equipment which causing rework cost on each project with level [min-max] = [1÷ 5]

 Impact level of others and Vietnam specific characteristic’s factors: the influence level of factors related to others and Vietnam specific characteristics which causing rework cost on each project with level [min- max] = [1÷ 5]

 Contract Amount: range [min-max] = [16.803.172.000 ÷

 Rework Percentage: range [min-max] = [1 ÷ 9]

 Rework Cost: Rework cost in comparison with total amount with range [min-max] = [168.031.720 ÷ 240.372.000.000]

Depend on collected data above, total 14 input data and 1 output data which data type are shown below:

Sign Data name Data Type

X6 Impact level of Investor’s factors Interger

X7 Impact level of Contractor/Subcontractor’s factors

X8 Impact level of supervisor’s factors Interger

X9 Impact level of Design Issue’s factors Interger

X10 Impact level of environment and external factors

X11 Impact level of labour and equipment’s factors Interger

X12 Impact level of others and Vietnam specific characteristic’s factors

According to Aksu, Güzeller et al 2019 [29], When raw data is used, multi- dimensional data sets are used, which leads to a number of issues, including lengthier processing times The normalization process of the raw input play an important role for reducing the size of the input field, noise reduction and feature extraction The normalization of data, which scales the data to the same range, effectively reduces bias in the artificial neural network Normalization of data accelerates the learning process for features covered in the same scale Such conveniences, many normalization method, among them one of the most popular is min-max method

ANN Modelling result

After setting up succesfully ANN model, proceed to run and the ANN Modelling result is shown as below:

Table 5.4 ANN Model result table

Figure 5.3 Comparison chart between Y and Y’

Figure 5.4 Correlation chart between Y and Y’ from ANN Model

Comment: According to ANN Model result parameter from Rapidminer Studio software, root mean square error (RMSE) value is 0.02 corresponding to 4,972,111,086 VND and mean absolute error (MAE) value is 0.014 corresponding to 3,530,887,276 VND both of value are considered as acceptable because those value are relative reflection for large scale projects in Vietnam Correlation coefficient (R) value is 0.995, which is relatively high, it is considered as a good linear correlation between the actual value and the predicted value of neurons value In addition, the value of mean absolute percentage error (MAPE) is 26.06% which is acceptable error percentage for ANN model In addtion, the collected data for rework percentage were not too correct with actual completion project such as estimated percentage number and not clear distinction rework part.

ANN Modelling comparison

Applying same dataset for analyzing Linear Regression model, Classification and regression trees (CART) model and Support Vector Machine (SVM) model The result of four criteria would be used for comparing among models consisting of correlation coefficient (R), mean absolute percentage error (MAPE), root mean square error (RMSE), and mean absolute error (MAE) The result are shown as table below:

Table 5.5 Comparisons of machine learning models table

Following to table 5.5, root mean square error (RMSE) and mean absolute error (MAE) of LR Model are value 0.034 (8,334,966,642 VND) and 0.023 (5,692,722,990 VND) RMSE and MAE of CART Model are value 0.040 (9,776,190,451 VND) and 0.024 (5,932,926,959 VND) RMSE and MAE of SVM Model are value 0.139 (33,556,383,311 VND) and 0.073 (17,702,921,404 VND)

Figure 5.5 Analysis result from Linear Regression model

Figure 5.6 Correlation chart between Y and Y’ from LR Model

Figure 5.7 Analysis result from CART model

Figure 5.8 Correlation chart between Y and Y’ from CART Model

Figure 5.9 Analysis result from SVM model

Figure 5.10 Correlation chart between Y and Y’ from SVM Model

Figure 5.11 Model correlation coefficient comparison chart

According to Figure 5.10, ANN model showed the highest value of correlation coefficient (R) with 0.995, then CART model with value 0.989, LR model is 0.989 and the lowest value is SVM model with 0.806 As conclusion, ANN model would be the most reliable model in comparison with others

Figure 5.12 Model MAPE comparison chart

According to Figure 5.11, MAPE describe the different error percentage within predicted value MAPE value of ANN model is shown 26.06% and considered is acceptable for range of collected data CART model MAPE value with lowest value 21.99% is better than LR model (34.85%) and SVM model (84.96%)

Figure 5.13 Model MAE and RMSE comparison chart

According to Figure 5.12, Both of value of root mean square error (RMSE), and mean absolute error (MAE) are defined as measures the mean error of the model compared to actual data and the lower value is, the better model is ANN model resulted with the lowest value of RMSE and MAE: 0.020 (4,972,111,086 VND) and 0.014

Mean absolute percentage error (MAPE) (%)

(3,530,887,276 VND) which is considered the best mean error within other models such as RMSE and MAE of LR Model are value 0.034 (8,334,966,642 VND) and 0.023 (5,692,722,990 VND) RMSE and MAE of CART Model are value 0.040 (9,776,190,451 VND) and 0.024 (5,932,926,959 VND) RMSE and MAE of SVM Model are value 0.139 (33,556,383,311 VND) and 0.073 (17,702,921,404 VND).

Chapter 5 Summary

ANN Model was set up and processed by Rapidminer Studio software 86 project information with component factors surveyed data were collected and created into 14 input data and 1 output data The result of ANN Model has been shown above with acceptable parameter about correlation coefficient (R), mean absolute percentage error (MAPE), root mean square error (RMSE), and mean absolute error (MAE) Additionally, in comparison with LR, CART and SVM model, ANN model has been determined as one of the best model for prediction value and applying on many life fields Besides, this research was recorded some limitation about collected data, but most of parameter have been showned on the acceptable results.

CONCLUSION AND RECOMMENDATION

Conclusion

Through this research project, factor causing rework were listed up and surveyed from respondents from construction field to describe the most factors impact on rework cost on both of level: occurrence and influence Research could help parties on construction field have had an overall look rework factors causing poor quality, extended construction schedule, cost overrun As a result, construction parties will have suitable plan and method for reducing impacts on rework Also on this research, factor components were described and determined impact of factors affecting rework costs in construction projects and ANN Model were created and estimated the rework cost on construction project from Rapidminer Studio 9.5 software

Pursuant to previous researches and opinion from experts, total of 46 factors on 7 main groups which were used for mass survey Mass survey method was applied by hardcopy directly and email indirectly, then 125 suitable questionnaire were collected from engineer and staff related to construction field The results for working experience

> 10 years is 47,2% and project type for civil project is 72,8 %, thus this questionnaire was considered appropriate and close with actual construction situation After getting data, collective data were treated and applied SPSS Statistic 25 software for data analysis Cronbach’s Alpha was used for reliability analysis of scale with occurrence and influence level, Cronbach’s Alpha coefficient for Occurrence & Influence level of each group have the value > 0,7 As the result, occurrence and influence level scale applied for this research is reliable Following to EFA progress with Extrection Method PCA and rotation method is Varimax, total of 43 factors were divided into 07 component group including others and Vietnam specific characteristics, Design Issue, Contractor/Subcontractor, Investor, Labour and Equipment, Supervisory, Environment and external impact

From the available data sources of project information such as contract agreement, addendum, which were collected during working period at bidding department and construction site GS E&C, were used on ANN Model However, data information related to rework cost percentage and factors impact on rework from SPSS Analysis were not available So one more collective data form was created and sent to engineer who experience on list of project for getting enough data All of collected data were used for setting ANN Data on Rapidminer Studio software ANN design structure including 3 layers: input layer (14 neurons), Hidden layer (4 neurons) and output layer (1 neuron) with Sigmoid function As a consequence, ANN prediction model for rework cost with the highest correlation coefficient value so ANN Model is more reliable than LR, CART, SVM model.

Recommendation

According to this research, factors impact on rework cost were determined in Vietnam construction project by surveying questionnaire Furthermore, this research create prediction model for rework cost from actual data which can be material for construction parties refer and estimate for making plan and method to reduce rework cost on projects Investor is considered as the most important role on occurance level and Contractor is the most impact group on influence level due to questionnaire results Depend on ANN model, rework cost percentage have to be controlled under 9% for each of project With large scale projects, the percentage of rework cost could be range from 1-2% which are seemed to be proper amount

The result of this research can be used as a reference and foundation for next research applying machine learning for rework cost problem

Investor can use result data for improving construction management procedure for minimizing rework influence and predict rework cost to make budget for initial stage of construction project

Using this research result can help Contractor avoid problem causing rework during construction period, improve manpower quality and estimate risk causing from external factors Moreover, prediction number for rework cost will be applied for contractor on bidding stage for having a suitable spare cost for rework cost

According to detail explanation, Design/ Supervisory consultant will have an management From those information, consultant parties will have good plan for reducing mistake and guarantee for construction quality, cost and time schedule

Stakeholder can refer this graduation thesis and model results as knowledge background for investment and cooperation with other parties In addition, those are good reference for stakeholder decicision on business target All of information can become a method for coordination evaluation by stakeholders.

Research limitation

Survey range was within Thu Thiem Zeit River project and GS E&C so it is considered as less variety for questionnaire

The “rework” meaning is actually not clear with most of people and just a few previous research described for rework issue

The questionnaire form is too long and quite complicated, the respondents were confused, so the results were not so accurate and objective

Because of confidential agreement of completed project, Engineer who joined completed project could not reveal exactly value for each of additional cost So the rework cost which received from respondents were almost percentage estimation, the exact value could not be provided following contract condition Moreover, some projects completed more than 5 years ago, so engineer in charge of those projects could remember rework value in details and they just provided an approximate number

Model function on Rapidminer software is seemed to be difficult for understanding and confusal for selecting function

Collected data which inserted on software need to take long time for refining in order to get good result

Next research target

“Rework” is the general mean which this research used for analysing, in the next research rework shall be researched on every detail of construction field such as rework cost related to communication issue, design issue, management issue,

Deep Leanring will be applied for the next model for rework cost analysis on Rapidminer Studio software

Questionnaire form need to be contribute with wider range for getting variety response

Widen research direction about other factors causing cost overrun on Vietnam construction project in order to analyze and apply machine learing for estimation which will support construction engineer reinforce their work knowledge.

[1] P E D Love, Z Irani, and D J Edwards, "A rework reduction model for construction projects," IEEE Transactions on Engineering Management, vol 51, no 4, pp 426-440, 2004, doi: 10.1109/TEM.2004.835092

[2] S Han, P Love, and F Peủa-Mora, "A system dynamics model for assessing the impacts of design errors in construction projects," Mathematical and Computer

Modelling, vol 57, no 9, pp 2044-2053, 2013/05/01/ 2013, doi: https://doi.org/10.1016/j.mcm.2011.06.039

[3] R Trach, Y Trach, and M Lendo-Siwicka, "Using ANN to Predict the Impact of Communication Factors on the Rework Cost in Construction Projects," Energies, vol 14, p 4376, 07/20 2021, doi: 10.3390/en14144376

[4] R Trach, M Lendo-Siwicka, K Pawluk, and M Połoński, "Analysis of direct rework costs in Ukrainian construction," Archives of Civil Engineering, vol 67, pp 397-411, 05/19 2021, doi: 10.24425/ace.2021.137175

[5] L Le-Hoai, Y Lee, and J.-Y Lee, "Delay and Cost Overruns in Vietnam Large Construction Projects: A Comparison with Other Selected Countries," KSCE

Journal of Civil Engineering, vol 12, pp 367-377, 11/01 2008, doi:

[6] T P K Toan, "Các nhân tố ảnh hưởng đến vượt dự toán trong các dự án xây dựng hạ tầng kỹ thuật tại thành phố Hồ Chí Minh Factors affecting cost overrun of infrastructure projects in Ho Chi Minh City," Tạp chí khoa học Đại học Mở Thành phố Hồ Chí Minh - kinh tế và quản trị kinh doanh, vol 16, no 3, pp 165-179,

[7] T H Tuấn, "Các nhân tố ảnh hưởng đến chi phí và thời gian hoàn thành dự án trong giai đoạn thi công trường hợp nghiên cứu trên địa bàn Thành phố Cần Thơ,"

Tạp chí Khoa học Đại học Cần Thơ, no 30, pp 26-33, 05/06 2014 [Online]

Available: https://ctujsvn.ctu.edu.vn/index.php/ctujsvn/article/view/1784

[8] M P Truong and V T Cao, "Factors affecting "re-work" in the design stage of construction projects," Tạp chí Xây Dựng, no 2021, 12/8/2021 2021

[9] National Assembly, “Law about amendments to Construction Law”, No

[10] Vietnam Government, “Decree about management of construction investment cost”, No 10/2021/ND-CP, February 9, 2021

[11] J M Dougherty, N Hughes, and J G Jr, "The impact of rework on construction and some practical remedies," AACE International Transactions, vol 1, pp 448-

[12] A Pudke, M D Milin, and N Shelke, "Factors affecting rework cost in construction," International research journal of engineering and technology (irjet), vol 7, no 8, pp 1017-1020, 2020

[13] W Pewdum, T Rujirayanyong, and V Sooksatra, "Forecasting final budget and duration of highway construction projects," Engineering, Construction and

Architectural Management, vol 16, pp 544-557, 11/06 2009, doi:

[14] A Al-Janabi, M Abdel-Monem, and K Eldash, "Factors causing rework and their impact on projects’ performance in Egypt," Journal of civil engineering and management, vol 26, pp 666-689, 08/21 2020, doi: 10.3846/jcem.2020.12916

[15] T Q Vu, C P Pham, T A Nguyen, P T Nguyen, P T Phan, and Q L H T T Nguyen, "Factors Influencing Cost Overruns in Construction Projects of

International Contractors in Vietnam," The Journal of Asian Finance, Economics and Business, vol 7, no 9, pp 389-400, 09/01 2020, doi:

[16] N Forcada, M Gangolells, M Casals, and M Macarulla, "Factors Affecting

Rework Costs in Construction," Journal of Construction Engineering and

Management, vol 143, p 04017032, 03/21 2017, doi: 10.1061/(ASCE)CO.1943-

[17] A Al-Janabi and M Abdel-Monem, "Review study for rework causes in construction industry," Al-Azhar University Civil Engineering Research Magazine

[18] H A Bình, "Những yếu tố quan trọng ảnh hưởng tới sự tăng mức đầu tư trong giai đoạn thi công công trình giao thông ở Việt Nam," Luận văn thạc sỹ, Ngành Công Nghệ & Quản Lý Xây Dựng, Trường Đại học Bách Khoa Tp.HCM, Trường Đại học Bách Khoa Tp.HCM, 2009

[19] J F Hair, W C Black, B J Babin, and R E Anderson, Multivariate Data

[20] N M N Chu and T Hoàng, - Phân tích dữ liệu nghiên cứu với SPSS Tp Hồ Chí Minh: Nhà xuất bản Hồng Đức, 2008

[21] L D Hai, "Áp dụng phương pháp phân tích nhân tố khám phá trong việc nghiên cứu các nhân tố ảnh hưởng đến động lực làm việc của người lao động trực tiếp tại tập đoàn Sentec Việt Nam," Tạp chí Khoa học và Lâm nghiệp, no 2018, 2018 [22] N D Tho, Giáo Trình Phương Pháp Nghiên Cứu Khoa Học Trong Kinh Doanh, 2 ed NXB Tài Chính, 2014, p 676

[23] J.-S Chou, D.-B Fleshman, and D.-N Truong, "Comparison of machine learning models for early estimates of real estate price," Journal of Housing and the Built

[24] X Ge, "Brief Introduction to Artificial Neural Networks," Culture Sciences de l’Ingénieur, ENS Paris-Saclay, France 7/2022 2022 [Online] Available: https://eduscol.education.fr/sti/sites/eduscol.education.fr.sti/files/ressources/pedago giques/14500/14500-brief-introduction-to-artificial-neural-networks-ensps.pdf [25] M Uzair and N Jamil, "Effects of Hidden Layers on the Efficiency of Neural networks," in 2020 IEEE 23rd International Multitopic Conference (INMIC), 5-7 Nov 2020 2020, pp 1-6, doi: 10.1109/INMIC50486.2020.9318195

[26] T Vujicic, T Matijevic, J Ljucovic, A Balota, and Z Sevarac, "Comparative Analysis of Methods for Determining Number of Hidden Neurons in Artificial Neural Network," Central European Conference on Information and Intelligent

Systems;Varazdin : Faculty of Organization and Informatics Varazdin: pp: 219-

[27] K Sheela and S N Deepa, "Review on Methods to Fix Number of Hidden

Neurons in Neural Networks," Mathematical Problems in Engineering, vol 2013, 01/01 2013, doi: 10.1155/2013/425740

[28] V Kotu and B Deshpande, "Chapter 4 - Classification," in Data Science (Second

Edition), V Kotu and B Deshpande Eds.: Morgan Kaufmann, 2019, pp 65-163

[29] G Aksu, C Güzeller, and T Eser, "The Effect of the Normalization Method Used in Different Sample Sizes on the Success of Artificial Neural Network Model,"

International Journal of Assessment Tools in Education, vol 6, pp 170-192, 07/01

[30] N Dang-Trinh, P Duc-Thang, T Cuong, and T Duc-Hoc, "Machine learning models for estimating preliminary factory construction cost: case study in Southern Vietnam," International Journal of Construction Management, vol 23, pp 1-9, 08/05 2022, doi: 10.1080/15623599.2022.2106043

[31] L H Q Phong, T D Hoc, N N Thuy, and P Q Tram, "Estimation the cost of insdustrial construction factory based on artifical neural network (ANN)," Tạp chí

ANALYZE THE IMPACT OF FACTORS AFFECTING REWORK COSTS

IN CONSTRUCTION PROJECTS USING ANN

Currently, I am student from Master Course- International Program Intake 2022, Construction Management of HCMC University of Technology- HCMC Vietnam

National University Now, I am carrying out graduation project with article: “Analyze the impact of factors affecting rework costs in construction projects using ANN’’

The purpose of this survey is to collect a database to evaluate and analyse the impact of "rework" costs in construction projects in Vietnam To get accurate information, we hope you will spend some valuable time to contribute your opinions by answering the following questions

The collected information will only be used for the purpose of the official graduation project and will not affect you or your company at all I commit that this information will be absolutely confidential

Thank you very much, ladies and gentlemen!

1 Number of years of experience in the construction industry:

2 Your current work unit is:

☐ Project Director ☐ Construction Manager/ Deputy

☐ Leader/ Deputy ☐ Engineer/ Construction Employee

4 Number of project you have participated:

5 Most of the projects that you have participated in are of the project type:

☐ Civil Project ☐ Bridge, Road/ Irrigation

6 Average scale of projects that you have participated in:

7 In the project you have participated in, the "rework" items had impact on many aspects of the project.:

8 "Rework" costs account for what percentage (%) of the total estimated costs in the projects that you have participated in?:

II PART 2: FACTORS IMPACT ON “REWORK” COST OF

In construction projects, the most important factors to evaluate the level of effectiveness include Progress - Cost - Quality Cost is the most powerful factor and causes many impacts on progress and quality Currently, many construction projects are completed with budgets exceeding the investment budget, which leads to a lot of controversy and loss of trust from users There are many causes of capital increases in projects and among them the costs of "rework" construction items are very concerning

The purpose of this graduation project is to identify and analyze the factors affecting the "rework" costs of Vietnamese projects, thereby evaluating and discussing the occurrence and influence in these projects

Corresponding to each question is a cause that can affect the increase in rework costs of construction projects with 5 levels of occurrence and 5 levels of influence of factors Respondents will choose the corresponding levels in the scale table:

Please mark an X in the box corresponding to your answer:

No The factors affecting the

Poor communication and connection with design parts/contractors

2 Decision making at the wrong time or late ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

3 Plans, designs, and adds scope of work changes ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

Lack of experience and understanding of the design and construction

II Factors related to Design

1 Incomplete design documents at the bidding stage ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

Too complicated design details, leading to difficult construction

Unclear and inconsistent communication and connection between members between design departments

5 Design details and structure missing in items ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

Inconsistent and incompliant design with standards and regulations

7 Vague and unsystematic design documents ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

III Factors related to Supervision Consultant

2 Incomprehensive clearly project design documents ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

3 Poor organization and supervision of project work ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

Inexperienced understanding regulations and information related to construction in

5 Inspection and control of design errors delay ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

IV Factors related to Contractors

1 Insufficient financial capacity to meet project requirements ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

2 Incompliant with design technical requirements ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

Poor communication and connection with project managers and supervisor

4 Poor quality construction materials choosing ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

5 Omissions, skipping steps in construction methods ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

7 Inefficient protection for completed items ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

Lack of connection and unity with the Investor and consultants

V Factors related to the environment and external factors

Protests and disputes over compensation and clearance issues

4 Complicated geological situation than the survey ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

VI Factors related to labour and equipment

1 Skilled workers shortage to complete the work ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

2 Inefficient safety guideline workers during construction ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

3 Ineffective training and developing skills for workers ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

4 Outdated machinery and construction technology ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

5 Broken and poor quality construction equipment ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

Changes in laws and government decisions led to many changes in construction

2 Outdated design due to long construction period ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

Obstacles in permit procedures related to other technical infrastructure projects

Errors or omissions in terms and scope of work in the contract

Technical measures and marketing technology that are too new

6 Ambiguous payment conditions between parties ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

Quality construction materials shortage in some localities in

9 Law ovelaps law in the

10 Low disbursement of capital sources ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐ ☐

III PART 3: PERCENTAGE OF COST INCREASE DUE TO

REWORK WORKS COMPARED TO ESTIMATE IN

According to you, the reasons leading to the above "rework" can increase the cost of construction by what percentage? (Please provide the percentage in the blank space) ……… (%)

* If you are interested in the survey, you can provide additional personal information to facilitate contact and provide a summary of the results related to this survey:

Once again, thank you very much!

For any information required, please contact:

Truong Cong Huy – Student K2022 International Master Program in Construction Management- HCMC University of Technology

Email: tchuy.sdh221@hcmut.edu.vn

APPENDIX II: PROJECT INFORMATION SURVEY FORM

II Impact level of factor groups on project

Corresponding to each question is impact level of particular componet group which is actual influence on completed construction projects with 5 levels of influence Respondents will choose the corresponding levels in the scale table:

No Group of factors impact on rework 1 2 3 4 5

1 Impact level of Investor’s factors ☐ ☐ ☐ ☐ ☐

2 Impact level of Contractor/Subcontractor’s factors ☐ ☐ ☐ ☐ ☐

3 Impact level of supervisor’s factors ☐ ☐ ☐ ☐ ☐

4 Impact level of Design Issue’s factors ☐ ☐ ☐ ☐ ☐

5 Impact level of environment and external factors ☐ ☐ ☐ ☐ ☐

6 Impact level of labour and equipment’s factors ☐ ☐ ☐ ☐ ☐

7 Impact level of others and Vietnam specific characteristic’s factors ☐ ☐ ☐ ☐ ☐

III Percentage of rework cost on project

According to you, how many percent "rework" cost occured on project? (Please provide the percentage in the blank space) ……… (%)

APPENDIX III: One-way ANOVA testing result for occurance level

Robust Tests of Equality of Means b,c

APPENDIX IV: One-way ANOVA testing result for influence level

Robust Tests of Equality of Means b,c,d

APPENDIX V: Exploratory factor analysis (EFA) rotated matrix table

Extraction Method: Principal Component Analysis

Rotation Method: Varimax with Kaiser Normalization a Rotation converged in 6 iterations

APPENDIX VI: DATASET APPLYING ON RAPIDMIDER STUDIO

APPENDIX VII: ANN MODEL RESULT

Node 1 Node 2 Node 3 Node 4 Threshold