A Feature Selection-Based K-NN Model for Fast Software Defect Prediction

Software Defect Prediction (SDP) is an advanced technological method of predicting software defects in the software development life cycle. Various research works have been previously being done on SDP but the performance of these methods varied from several datasets, hence, making them inconsistent for SDP in the unknown software project. But the hybrid technique using feature selection enabled with machine learning for SDP can be very efficient as it takes the advantage of various methods to come up with better prediction accuracy for a given dataset when compared with an individual classifier. The major issues with individual ML-based models for SDP are the long detection time, vulnerability of the software project, and high dimensionality of the feature parameters. Therefore, this study proposes a hybrid model using a feature selection enabled Extreme Gradient Boost (XGB) classifier to address these mentioned challenges. The cleaned NASA MDP datasets were used for the implementation of the proposed model, and various performance metrics like F-score, accuracy, and MCC were used to reveal the performance of the model. The results of the proposed model when compared with state-of-the-art methods without feature selection perform better in terms of the metrics used. The results reveal that the proposed model outperformed all other prediction techniques.