Software Fault Prediction using Wrapper based Ant Colony Optimization Algorithm for Feature Selection

Abstract

Feature selection is both important and difficult part in classification technology. It is used to reduce the dimensionality of dataset and remove unwanted features. In this paper, we are going to discuss about a supervised feature selection method based on ant colony optimization for software fault prediction. Software fault prediction helps to identify software fault at an early stage. We have used KNN, Naive Bayes and Decision Tree classifiers. We designed a fitness function and a two-step pheromone update rule is applied for effective elimination of duplicate features. This algorithm is inspired from real ants that search for the shortest path to the food source depending upon the concentration of the pheromone. We have used 12 different datasets and compared them using fitness plots. Each and every graph represents the effectiveness of ant colony optimization along with different classifiers. We have also made a table that represents the accuracy of the prediction for all the different classifiers while using the algorithm.