A Novel Feature Selection Method based on an Integrated Data Envelopment Analysis and Entropy Model

Abstract

Data mining is a one of the growing sciences in the world that can play a competitive advantages rule in many firms. Data mining algorithms based on their functions can be divided in four categories; Classification, Feature selection, Assassination rules and Clustering. One of the most important of these functions is feature selection which has been increasingly developed and many researchers provide variety of algorithms to deal with this function in recent years. Feature selection algorithms mostly used for obtaining more precise and strong machine learning algorithms along with reducing the computation time. Another growing science is Multiple Criteria Decision Making techniques witch it also has variety of methods. In this paper, we use both Data Envelopment Analysis which is a useful technique for determining the efficiency of decision-making units and Entropy method which its function is weighting the criteria to selecting the appropriate features. Hence, our novel integrated method has been analyzed by implementing in a testing environment and we apply it on three datasets of UCI's datasets, so the result showed our innovated approach has comparable accuracy with the other feature selections algorithms.