Feature Selection Using Diploid Genetic Algorithm

Classification is an important task in Machine Learning. Often datasets used for such problems have a large number of features where only a few may be actually useful for this task. Feature Selection is the process where we aim to remove irrelevant features in order to improve our performance. This improved performance could be achieved with an increase in accuracy or by minimizing number of features selected for the task of classification, most Feature Selection algorithms aims at only one of these objectives in their approach. This paper presents the use of Diploid Genetic Algorithm (DGA) on multi-objective optimization of a classification problem for feature selection. The task is to develop a model for solving a Subset Sum Problem using DGA and applying the solution of this problem in order to accomplish the goal of multi-objective optimization by maximizing accuracy using minimum number of features. The model has been applied to publicly available datasets and the results shown are encouraging. This work establishes the veracity of DGA in feature selection.