&KX\rQQJjQK .+2$+Ӑ&0È<7Ë1+ MmVӕ 60.48.01.01
/8Ұ19Ă17+Ҥ&6Ƭ
73+Ӗ&+Ë0,1+WKiQJQăP
Trang 2&Ð1*75Î1+ĈѬӦ&+2¬17+¬1+7Ҥ, 75ѬӠ1*ĈҤ,+Ӑ&%È&+.+2$±Ĉ+4*-HCM &iQEӝKѭӟQJGүQNKRDKӑF: *676&DR+RjQJ7Uө
769}7Kӏ1JӑF&KkX
&iQEӝFKҩPQKұQ[pW
&iQEӝFKҩPQKұQ[pW
/XұQYăQWKҥFVƭÿѭӧFEҧRYӋWҥL7UѭӡQJĈҥLKӑF%iFK.KRDĈ+4*7S+&0QJj\WKiQJQăP 7KjQKSKҫQ+ӝLÿӗQJÿiQKJLi OXұQYăQWKҥFVƭJӗP *KLU}KӑWrQKӑFKjPKӑFYӏFӫD+ӝLÿӗQJFKҩPEҧRYӋOXұQYăQWKҥFVƭ 1
Trang 375ѬӠ1*ĈҤ,+Ӑ&%È&+KHOA
&Ӝ1*+Ñ$;+Ӝ,&+Ӫ1*+Ƭ$9,ӊ7NAM ĈӝFOұS- 7ӵGR- +ҥQKSK~F
1+,ӊ09Ө/8Ұ19Ă17+Ҥ&6Ƭ
+ӑWrQKӑFYLrQ 3+Ҥ07581*6Ѫ1 MSHV: 1670016 1Jj\WKiQJQăPVLQK 24/11/1993 1ѫLVLQK %uQK'ѭѫQJ Chuyên ngành: KRD+ӑF0i\7tQK 0mVӕ : 60.48.01.01
III 1*¬<*,$21+,ӊ09Ө : 13/08/2018
IV 1*¬<+2¬17+¬1+1+,ӊ09Ө: 02/06/2019
V &È1%Ӝ+ѬӞ1*'Ү1 *KLU}KӑFKjPKӑFYӏKӑWrQ
GS TS &DR+RjQJ7Uө TS 9}7Kӏ1JӑF&KkX
Tp HCM, QJj\WKiQJQăP
&È1%Ӝ+ѬӞ1*'Ү1 +ӑWrQYjFKӳNê
&+Ӫ1+,ӊ0%Ӝ0Ð1Ĉ¬27Ҥ2 +ӑWrQYjFKӳNê
GS TS &DR+RjQJ7UөTS 9}7Kӏ1JӑF&KkX
75ѬӢ1*.+2$«««« +ӑWrQYjFKӳNê
Trang 4/Ӡ,&Ҧ0Ѫ1
ĈҫX WLrQ chúng tôi [LQ FKkQ WKjQK FҧP ѫQ ÿӃQ WKҫ\ &DR +RjQJ 7Uө Yj F{ 9} 7Kӏ1JӑF&KkXNKRD.KRD+ӑFYj.ӻ7KXұW0i\7tQKÿmWұQWuQKKѭӟQJGүQchúng tôi WӯNKkXOӵDFKӑQÿӅWjLWKHRG}LVӱDFKӳDKRjQWKLӋQFKRFK~QJW{LVXӕWtrong quá trình WKӵFKLӋQFNJQJQKѭ WKHRViWWUDRÿәLYӅQKLӅXYҩQÿӅPӣPDQJFKRchúng tôi QKLӅXKѭӟQJÿLPӟLJL~Schúng tôi KRjQWKLӋQWӯêWѭӣQJÿӃQKLӋQWKӵFOtOXұQFNJQJQKѭcách trình bày
Chúng tôi FKkQWKjQKELӃWѫQVӵWұQWuQKGҥ\GӛJL~SÿӥFӫDFiFWKҫ\F{WURQJNKRD.KRDKӑFYj.ӻWKXұW0i\WtQKÿmWUX\ӅQÿҥWQKӳQJNLQKQJKLӋPNLӃQWKӭFQKӳQJEjLJLҧQJY{FQJVkXVҳFYjKӳXtFKPjQJj\K{PQD\ÿmÿѭӧFYұQGөQJWURQJÿӅWjL &XӕLFQJchúng tôi JӱLOӡLFҧPѫQÿӃQJLDÿuQKEҥQEqQKӳQJQJѭӡLÿmTXDQWkPÿӝQJYLrQJL~SÿӥFҧYӅWKӇFKҩWOүQWLQKWKҫQÿӇFK~QJW{L FyÿӫQJKӏOӵFVӭFNKӓHKRjQWKjQKWӕWÿѭӧFÿӅWjLOXұQYăQ
9ӟLOzQJELӃWѫQFKkQWKjQKchúng tôi [LQJӱLOӡLFK~FVӭFNKӓHOӡLELӃWѫQYjQKӳQJÿLӅX WӕWÿҽSQKҩWWӟLFiFWKҫ\F{ӣNKRD.KRDKӑFYj.ӻWKXұW 0i\WtQKWUѭӡQJĈҥL+ӑF%iFK.KRDWKjQKSKӕ+ӗ&Kt0LQK
7UkQWUӑQJ
Trang 57Ï07Ҳ7/8Ұ19Ă1
9ӟLVӵSKiWWULӇQFӫDPҥQJ[mKӝLQJj\FjQJFyQKLӅXWK{QJWLQFKӫÿӅÿѭӧFFKLDVҿEjQOXұQYjWKXK~WQKLӅXQJѭӡLGQJKѫQ&iFEjLWRiQSKiWKLӋQSKkQWtFKGӵEiRFKRFiFFKӫÿӅQyQJÿѭӧFTXDQWkPQJKLrQFӭXGRFyêQJKƭDWKӵFWLӉQFDRWURQJFiFlƭQKYӵFӭQJGөQJNKiFQKDXQKѭWLӃSWKӏTXҧQJEiQӝLGXQJ
4XDQJKLrQFӭXEjLWRiQYjFiFF{QJWUuQKOLrQTXDQFK~QJW{LÿmQҳPEҳWNKҧRViWÿѭӧFEjLWRiQGӵEiRFKӫÿӅQyQJYӅWKӵFWUҥQJWKiFKWKӭFFNJQJQKѭFiFÿһFWUѭQJYjP{KuQKGӵEiRÿѭӧFVӱGөQJFKREjLWRiQ7UrQFѫVӣÿyFK~QJW{LÿmÿӅ[XҩWNӃWKӧSFiFQKyPÿһFWUѭQJYj[k\GӵQJSKѭѫQJSKiSKuQKWKjQKPүXGӳOLӋXkP-GѭѫQJFKREjLWRiQGӵEiRFKӫÿӅQyQJ6DXÿyFK~QJW{LJLҧLTX\ӃWEjLWRiQQKѭEjLWRiQSKkQOӟSQKӏSKkQYӟLKѭӟQJWLӃSFұQ KӑFPi\GQJFiFJLҧLWKXұWKӑFFyJLiPViW 9ӟLFiFÿӅ[XҩWWUrQFKROXұQYăQFK~QJW{Lÿm[k\GӵQJEjLWRiQSKiWWULӇQJLҧLSKiSYjWKӵFKLӋQFiFWKtQJKLӋPÿiQKJLiPӝWFiFKFKһWFKӁÿҫ\ÿӫQKҵPWҥRPӝWFѫVӣVRViQKFKRFiFF{QJWUuQKVDX&iFNӃWTXҧWKtQJKLӋP NKҧTXDQ ÿѭӧFFҧLWKLӋQYӟLFiFÿӅ[XҩWÿyQJJySFӫDOXұQYăQ, FKREjLWRiQGӵEiRFKӫÿӅQyQJ
Trang 6ASBTRACT
With the development of social networks, more and more information, topics are shared, discussed, and attracted a lot of users The problems of detecting, analyzing and predicting for hot topics are interested in research due to their high practical meaning in different application areas such as marketing and content promotion
Through studying the problem and related works, we have grasped and surveyed the hot topic prediction problem about the situation and challenges as well as the characteristics and predictive models used for the problem On that basis, we have proposed combining feature groups and developing methods to form positive and negative data samples for the problem of hot topic prediction Then we solve the problem as a binary classification problem with a machine learning approach, using supervised learning algorithms With the above suggestions for the dissertation, we have stated a problem, developed a solution, and conducted rigorous and complete evaluation experiments to create a comparative basis for the following works Experimental results were positive, improved with suggestions of the thesis, for the hot topic prediction problem
Trang 7/Ӡ,&$0Ĉ2$1
Chúng t{LFDPÿRDQUҵQJQJRҥLWUӯFiFNӃWTXҧWKDPNKҧRWӯFiFQJXӗQNKiFFyghi rõ trong WjLOLӋXWKDPNKҧRSKөOөFWKuFiFF{QJYLӋFWUuQKEj\WURQJOXұQYăQWӕWQJKLӋS Qj\ ÿӅX FKR FKtQKW{LWKӵF KLӋQYj FKѭD Fy SKҫQ QӝLGXQJQjR FӫD OXұQYăQ ÿѭӧF[XҩWEҧQVӱGөQJ ӣWUѭӡQJKRһFQѫLNKiF1ӃXFyEҩWNǤVDLSKҥPQjRW{L[LQFKӏX KRjQ WRjQ WUiFK QKLӋP WUѭӟF %DQ &Kӫ 1KLӋP KRD Yj %DQ *LiP +LӋX 1Kj7UѭӡQJ
7S+ӗ Chí Minh, tháng 6 QăP9
+ӑFYLrQ WKӵFKLӋn
3K̩P7UXQJ6˯Q
Trang 83.1 PHÁT BIӆU BÀI TOÁN DӴ BÁO CHӪ Ĉӄ NÓNG 13
3.2 CÁC CÂU HӒI NGHIÊN CӬU 14
3.3 THÁCH THӬC 15
&KѭѫQJ 3+ѬѪ1*3+È3Ĉӄ XUҨT 16
4.1 TӘNG QUAN Vӄ +ѬӞNG TIӂP CҰN GIҦI QUYӂT BÀI TOÁN 16
4.2 ;È&ĈӎNH MҮ8'ѬѪ1*0ҮU ÂM 17
4.2.1 3KѭѫQJSKiSFKXҭn bӏ dӳ liӋu theo cӵFÿҥi toàn cөc 17
4.2.2 3KѭѫQJSKiSFKXҭn bӏ dӳ liӋXWKHRQJѭӥng sӕ OѭӧQJEjLÿăQJ 18
4.3 5Ò775Ë&+ĈҺ&75Ѭ1* 19
4.3.1 CiFÿһFWUѭQJVӱ dөng 19
Trang 94.4 HIӊN THӴC MÔ HÌNH DӴ BÁO 22
Trang 10DANH MӨC HÌNH
+uQK6ѫÿӗ tәng quát cӫa hӋ thӕng phát hiӋn và dӵ báo chӫ ÿӅ nóng trên mҥng xã hӝi 8Hình 2 Bài toán dӵ báo chӫ ÿӅ nóng 13+uQK+ѭӟng tiӃp cұn hӑc có giҧm sát vӟLFiFÿһFWUѭQJÿѭӧc rút trích cho bài toán dӵ báo chӫ ÿӅ nóng 16Hình 4 MүXkPYjGѭѫQJYӟLSKѭѫQJSKiSFKXҭn bӏ dӳ liӋu theo cӵFÿҥi toàn cөc 18Hình 5 Mүu âm và mүX GѭѫQJ YӟL SKѭѫQJ SKiS FKXҭn bӏ dӳ liӋX WKHR QJѭӥng sӕ OѭӧQJEjLÿăQJ 19+uQK6ѫÿӗ xây dӵng mô hình dӵ báo chӫ ÿӅ nóng 22Hình 7 BiӇXÿӗ ÿӝ ÿR)Fӫa thí nghiӋm vӟLSKѭѫQJSKiSFKXҭn bӏ dӳ liӋu theo cӵc ÿҥi toàn cөc 29Hình 8 biӇX ÿӗ ÿӝ ÿR ) Fӫa thí nghiӋm vӟi SKѭѫQJ SKiS FKXҭn bӏ dӳ liӋu theo QJѭӥng sӕ OѭӧQJEjLÿăQJ 30
Trang 11DANH MӨC BҦNG BIӆU
BҧQJ&iFÿһFWUѭQJVӱ dөng 19BҧQJ&iFSKѭѫQJSKiSSKkQOӟp sӱ dөng 24Bҧng 3 KӃt quҧ thí nghiӋm vӅ tӯQJÿһFWUѭQJWUrQEӝ dӳ liӋu vӟLSKѭѫQJSKiSFKXҭn bӏ dӳ liӋu theo cӵFÿҥi toàn cөc 25Bҧng 4 KӃt quҧ thí nghiӋm vӅ tӯQJÿһFWUѭQJWUrQEӝ dӳ liӋu vӟLSKѭѫQJSKiSFKXҭn bӏ dӳ liӋXWKHRQJѭӥng sӕ OѭӧQJEjLÿăQJ 25Bҧng 5 KӃt quҧ thí nghiӋm kӃt hӧS FiF ÿһF WUѭQJ WUrQ Eӝ dӳ liӋu vӟL SKѭѫQJ SKiSchuҭn bӏ dӳ liӋu theo cӵFÿҥi toàn cөc 26Bҧng 6 KӃt quҧ thí nghiӋm kӃt hӧS FiF ÿһF WUѭQJ WUrQ Eӝ dӳ liӋu vӟL SKѭѫQJ SKiSchuҭn bӏ dӳ liӋXWKHRQJѭӥng sӕ OѭӧQJEjLÿăQJ 27Bҧng 7 KӃt quҧ thí nghiӋm vӅ FiFSKѭѫQJSKiSSKkQOӟp trên bӝ dӳ liӋu vӟLSKѭѫQJpháp chuҭn bӏ dӳ liӋu theo cӵFÿҥi toàn cөc 27Bҧng 8 KӃt quҧ thí nghiӋm vӅ FiFSKѭѫQJSKiSSKkQOӟp trên bӝ dӳ liӋu vӟLSKѭѫQJpháp chuҭn bӏ dӳ liӋXWKHRQJѭӥng sӕ OѭӧQJEjLÿăQJ 28Bҧng 9 KӃt quҧ ÿӝ ÿR)Fӫa thí nghiӋm vӟLSKѭѫQJSKiSFKXҭn bӏ dӳ liӋu theo cӵc ÿҥi toàn cөc 29Bҧng 10 KӃt quҧ ÿӝ ÿR ) Fӫa thí nghiӋm vӟL SKѭѫQJ SKiS FKXҭn bӏ dӳ liӋu theo QJѭӥng sӕ OѭӧQJEjLÿăQJ 30
Trang 12DANH MӨC TӮ VIӂT TҲT
API Application Programming Interface
GBDT Gradient Boosting Decision Tree
PAA Piecewise aggregate approximation
Trang 13&KѭѫQJ GIӞI THIӊU
1.1 TӘNG QUAN
1Jj\QD\YӟLVӵSKiWWULӇQPҥQKPӁFӫD F{QJ QJKӋ [ӱOtOѭXWUӳ và WUX\ӅQ WҧLGӳOLӋXFiFKFRQQJѭӡLJLDRWLӃS và WuPNLӃPWK{QJWLQFNJQJWKD\ÿәL7KD\YuFKӍWLӃSQKұQWK{QJWLQPӝWFiFKWKөÿӝQJ, JLӡÿk\QJѭӡLGQJOҥLWUӣWUjQKQKkQWӕFKtQKWURQJYLӋFFKLDVҿWK{QJWLQYӟLQKXFҫXvà WӕFÿӝWѭѫQJWiFWăQJOrQQKLӅXOҫQVRYӟLFiFSKѭѫQJSKiSWUX\ӅQWKӕQJFNJ&iFmҥQJ[mKӝL UDÿӡLYj ÿDQJGҫQÿLYjRFXӝFVӕQJKҵQJQJj\FӫDFRQQJѭӡL&QJYӟLVӵSKiWWULӇQFӫDFiFWKLӃWEӏGLGӝQJÿһFELӋWOjÿLӋQWKRҥLWK{QJPLQKQJѭӡLGQJFyWKӇGӉGjQJÿăQJEjLYLӃWWUҥQJWKiLFiQKkQJӱLWLQ QKҳQ FKLD Vҿ WLQ WӭF Yj WKDP JLD WKҧR OXұQ FiF FKӫ ÿӅ WUrQ QKLӅX OƭQK YӵF NKiFnhau
9ӟL KjQJWӹQJѭӡLGQJPҥQJ[mKӝLWUӣWKjQKQѫLF{QJFөchính, ÿӇFRQQJѭӡLWKXQKұS Yj FKLD Vҿ WK{QJ WLQ 0ӝW Vӕ Yt Gө Fy WKӇ NӇ ÿӃQ QKѭ )DFHERRN và Twitter Facebook là PӝWPҥQJ[mKӝLYӟLVӕOѭӧQJQJѭӡLGQJKRҥWÿӝQJKjQJWKiQJOrQÿӃQFRQVӕ Wӹ1 Twitter là PӝW PҥQJ [m KӝL chia Vҿ, Fy WKӇ Fy NKRҧQJ EjL YLӃW ÿѭӧF ÿăQJ PӛL JLk\ Yj NKRҧQJ WULӋX PӛL QJj\2 7K{QJ WLQ ÿѭӧF FKLD Vҿ UҩW ÿDGҥQJ SKRQJSK~WӯQKӳQJ GzQJFұS QKұWWUҥQJ WKiLFi QKkQ êNLӃQ FҧP[~F NLQKQJKLӋPFKRÿӃQQKӳQJWLQWӭFPӟLQKҩWFNJQJQKѭQKӳQJFұSQKұWYjWKҧROXұQYӅQKӳQJVӵNLӋQQyQJÿDQJGLӉQUDWUrQWKӃJLӟL 0ҥQJ[mKӝL FzQFyWKӇJKLQKұQQKDQKFKyQJFiFEuQKOXұQSKҧQӭQJWKiLÿӝ, và FҧP[~FFӫDQJѭӡLÿӑFYӅFiFFKӫÿӅ Bên FҥQKÿy FiFFKӫÿӅQyQJFNJQJҧQKKѭӣQJQJѭӧFOҥLNK{QJQKӓÿӃQVX\QJKƭTX\ӃWÿӏQKFӫDQJѭӡLGQJWURQJFiFOƭQKYӵFQKѭÿҫXWѭWLӃSWKӏYLӋFOjP«
9ҩQÿӅNKDLWKiFWK{QJWLQKLӋXTXҧWӯPҥQJ[mKӝLWUӣQrQFӵFNǤFҫQWKLӃW Ta QKұQUDWLӅPQăQJNKDLWKiFWK{QJWLQWӯYLӋFSKiWKLӋQGLӉQGӏFKSKkQWtFKFiFFKӫÿӅÿѭӧFEjQOXұQWUrQPҥQJ[mKӝL 9tGөQKѭFiFFKӫÿӅFyWKӇFXQJFҩSQKLӅXWK{QJWLQKӳXtFKWURQJWLӃSWKӏVҧQSKҭPĈһF WtQKODQWӓD QKDQK FKyQJ FӫD FiF FKӫ ÿӅ QyQJWUrQ
1 https://zephoria.com/top-15-valuable-facebook-statistics/
2 http://www.internetlivestats.com/twitter-statistics/
Trang 14PҥQJ [m KӝL FNJQJ ÿѭӧF WұQ GөQJ WURQJ YLӋF [k\ GӵQJ WKѭѫQJ KLӋX và TXҧQJ Ei QӝLdung
'RÿyÿӇ ÿiS ӭQJQKXFҫX QҳPEҳWÿѭӧF FiF FKӫ ÿӅ QyQJFNJQJ QKѭQҳPEҳWÿѭӧF QJѭӡLGQJYjVӵWѭѫQJWiFFӫDKӑFiFEjLWRiQOLrQTXDQÿӃQSKiWKLӋQSKkQWtFKGӵ
EiRYӅFiFFKӫÿӅQKѭSKiWKLӋQFKӫÿӅQyQJ hot topic detection GӵEiRFKӫÿӅQyQJ(hot topic prediction ÿѭӧFÿһWUDYjTXDQWkPQJKLrQFӭX
&QJYӟLVӕOѭӧQJEjLÿăQJWLQQKҳQEuQKOXұQFNJQJQKѭVӕOѭӧQJQJѭӡLGQJNKәQJOӗ FiF WK{QJ WLQ WUrQ PҥQJ [m KӝL Fy WKӇ ODQ WӓD PӝW FiFK QKDQK FKyQJ VR YӟL FiFSKѭѫQJWLӋQWUX\ӅQWK{QJWUѭӟF ÿk\ %jLWRiQ[ӱ OtWK{QJWLQYӅFiFFKӫÿӅWKҧROXұn QKѭEjLWRiQ GӵEiR FKӫÿӅQyQJ ÿzLKӓLQKӳQJ\rXFҫXYӅKLӋXTXҧ YjWKӡLJLDQ[ӱOtNKҳW NKH Ngoài ra, ciF WK{QJ WLQ FKӫ ÿӅ Vӵ NLӋQ ÿѭӧF ÿăQJ WҧL WUrQ PҥQJ [m KӝLWKѭӡQJFy WtQK PӟLUҩWFұS QKұWVRYӟLFiF NrQK WK{QJWLQ NKiF QKѭEiR FKtWUX\ӅQWKӕQJĈһFÿLӇPQj\ NKLӃQFKRYLӋF[ӱOêFӫDFiFEjLWRiQOLrQTXDQ WUӣQrQNKyNKăQ và SKӭFWҥSKѫQ
3KiWKLӋQFKӫÿӅQyQJOjYҩQÿӅÿѭDUDÿѭӧFPӝWGDQK ViFKFiFFKӫÿӅQyQJWӯWұSKӧSFiFEjLÿăQJWK{QJÿLӋSWKҧROXұQFӫDQJѭӡLGQJWҥLWKӡLÿLӇPKLӋQWҥLӢEjLWRiQSKiWKLӋQFKӫÿӅQyQJKLӋQWҥLÿmFyQKLӅXF{QJWUuQKQJKLrQFӭXÿѭӧFF{QJEӕYӟLQKLӅXFiFKWLӃSFұQNKiFQKDXYjFKRNӃWTXҧUҩWNKҧTXDQQKѭKӋWKӕQJ7ZLWWHU1HZV
[1], TwitterNews+ [2] YӟLKѭӟQJWLӃSFұQJӑPFөPJLDWăQJ based FKRFiFFKӫÿӅ+LӋQQD\KӋWKӕQJ7ZLWWHU1HZVVDXNKLÿѭӧFWLQKFKӍQKYj
(Incremental-clustering-F{QJ EӕQăP [2] ÿҥWNӃWTXҧ SKiW KLӋQOrQÿӃQ WUrQ YӟLEӝGӳOLӋX(YHQW2012 [3] 1JRjLUDFzQFyPӝWVӕKѭӟQJWLӃSFұQNKiFGӵDWUrQÿӝWK~Yӏ FӫDWӯNKyD
(Term-interestingness-based ÿѭӧF iS GөQJ WURQJ FiF F{QJ WUuQK WLrX ELӇX QKѭ KӋ
WKӕQJ 7ZHYHQW [4] và Twitinfo [5] KѭӟQJ WLӃS FұQ GӵD WUrQ P{ KuQK KyD FKӫ ÿӅ
(Topic-modelling-based) [6]
1JѭӧFOҥLFiFF{QJWUuQKQJKLrQFӭXYӅGӵEiRFKӫÿӅQyQJOҥLFyQKLӅXNӃWTXҧUӡLUҥFWUrQQKLӅXEӝGӳOLӋXNKiFQKDXYjKҫXKӃWOjNK{QJÿѭӧFF{QJNKDL&iFNӃWTXҧQj\ NKLӃQ FKR FiF ÿӅ [XҩW KLӋQ WҥL cho bài WRiQ Gӵ EiR FKӫ ÿӅ QyQJ mang tính lý WKX\ӃWNKyVRViQK FK~QJW{LVӁWUuQKEj\NӻKѫQӣ&KѭѫQJ &{QJWUuQKOLrQTXDQ