A solution approach based on polyhedral conic functions for feature selection problems

In this study, an embedded method based on polyhedral conic functions has been proposed for feature selection problems, which is one of the important branches of machine learning. The polyhedral conic functions algorithm is a supervised classification algorithm developed for separating the sets which of convex hulls are intersect. It is reported on different studies in literature that PCF algorithm gives competitive classification accuracies. The purpose of this study is to develop an algorithm that includes polyhedral conic functions for feature selection problems. The classical algorithm based on the separation theory has been rearranged by defining a new objective function which includes a feature selection term and a new termination criteria. Thanks to the new objective function both feature selection and classification can be performed simultaneously. The proposed algorithm is applied to solve classification problems on some well known real world data sets and compared with other well known classifiers. Finally, an illustrative example is presented using a student performance data set.