Software Defect Prediction System using Machine Learning based Algorithms

Abstract

Measuring the performance, reliability or quality of a software simply describes the sequence of actions taken detecting bugs in a software product. The Bugs found during the development of software has made researchers develop different methods of bug prediction models. However, predicting the bugs in a concurrent software product reduces development time and cost. In this paper, experiments were conducted on public available bug prediction dataset which is a repository for most open source software. The Genetic algorithm was used to extract relevant features from the acquired datasets to eliminate the possibility of over-fitting. The extracted features are classified to defective or non-defective using random forest, decision tree and artificial neural network classification technique. Furthermore, the techniques were evaluated using accuracy, precision, recall and f-score. In completion of the conducted experiments, the random forest performs best among the algorithms in terms of accuracy, precision, and f-score with average score of 83.40%, 53.18%, and 52.04% respectively. Also, the results showed that neural network performs best in terms of recall with average score of 60% among the algorithms. Hence, the system helped software developers when developing a good quality software in order to check if the software system has a little or no defects before delivery to customers.