This paper presents an ensemble model for software defect prediction using method-level features of a Spring Framework open-source Java project called Broadleaf Commerce. The proposed model uses a combination of 3 machine learning algorithms such as random forest, support vector machine using RBF Kernel, and LightGBM to predict the likelihood of software defects in the project. The method-level features considered for defect prediction include method complexity, method calls, and method length. The proposed ensemble model achieves a high ROC curve of 0.853 in defect prediction, outperforming the individual machine learning algorithms. The study demonstrates the effectiveness of using method-level features and ensemble models for software defect prediction in open-source Java projects. The results of the study can help software developers to identify potential defects and take corrective actions to improve software quality.

Ensemble model, Software Defect Prediction, LightGBM, Random Forest, Software Quality