Neural Network based Software Defect Prediction using Genetic Algorithm and Particle Swarm Optimization

In the arena of software engineering, software defects prediction is one of the most attractive research topics. Here the main task is to predict if there is any bug in the software or not. For software testing, software defect detection is important for reducing the time and resources consumed. Accurate estimate of defect software prediction process enables effective discovery and identification of the defects. Such prediction methods are important for the big scale systems, where verification specialists need to focus their attention. In this paper, we proposed a method where the features are selected using Genetic Algorithm (GA). Secondly, make cluster of the selected features using Particle Swarm Optimization (PSO) and then train the model with different Neural Network (NN) methods such as: Feedforward Neural Network (FNN), Recurrent Neural Network (RNN), Artificial Neural Network (ANN), and Deep Neural Network (DNN) and finally calculate accuracy, sensitivity, specificity, precision, negative prediction value, F1 score, and Matthews correlation coefficient. We use five different datasets from NASA promise software engineering repository for our study. From our study, we get the best accuracy result using deep neural network. Experimental consequences show that proposed strategy is a decent technique to predict the software defects.