... following formula (Yang and Pedersen, 1997) ( , ) log( ( | )) log( ( ))i iI t c P t c P t= − 694Therefore, this method might perform badly when common terms are informative for classification. ... strongly prefer frequency information, e.g., DF. 4.3 Performances of Different FS Methods It is worth noting that learning parameters in WFO is very important for its good performance. We use 9-fold ... the terms with higher document frequency are more informative for classification. But sometimes this assumption does not make any sense, for example, the stop words (e.g., the, a, an) hold...