Software Change Prediction Using Ensemble Learning

Software change can significantly impact software quality and reliability, increasing costs, decreasing customer satisfaction, and even safety risks. Change prediction is a critical task of software engineering to identify major changes before they occur. This study focusses on various techniques for predicting changes in software using ensemble machine learning (ML) techniques. The approach involves using a set of static code metrics as features for change prediction and an ensemble of five ML algorithms, namely decision tree, random forest, K nearest neighbor, Logistic Regression, and support vector machine, to predict software changes. The ensemble is created using a stacking technique, where the predictions of the base classifiers are combined using a meta-classifier. The approach evaluated on dataset of changes from real world project and compared with five baseline methods. The results show the comparison of various approach which outperforms the baseline methods regarding several evaluation metrics, as recall, precision, and F1-score. The proposed work is an effective method for predicting software changes, and that ensemble ML techniques can be a powerful tool for improving the accuracy of change prediction.