Feature Selection in Data Mining using Chemical Reaction Optimization Algorithm

One of the characteristics of recent problems can be referred to the great number of features that have led to slowing down the classification systems, decreased efficiency and rising the costs of such systems. In recent years, feature selection problem has been investigated in data mining field when encountering data sets with many features. This study aimed to present new and optimized application in order to use metaheuristic algorithms in the feature selection problem in data sets in which through large number of useless features can decrease these features and their required time to implement various algorithms. In this study, a metaheuristic algorithm called Chemical Reaction Optimization Algorithm was used that is among the modern and most powerful evolutionary optimization techniques introduced in 2010. At the end, the proposed method was analyzed along with available standard data sets of UCI. The results indicated high efficiency based on two criteria of classification accuracy and small subset selection of features as the salient features at the same time.