Impact of Data Characteristics on Feature Selection Techniques Performance

Abstract

Feature selection is a step in knowledge discovery in databases which takes away most of the time of the entire process. Therefore, the effective implementation of feature selection significantly improves the overall process. This paper suggests examining data characteristics before applying feature selection and hypothesize that data characteristics significantly affect feature selection techniques performance. Our experimental comparison of five previously used feature selection techniques reveals significant difference in feature selection techniques performance when dealing with data sets of different characteristics