Abstract:Software fault prediction is a critical aspect of software engineering aimed at improving software quality and reliability. However, it faces significant challenges, including the class imbalance issue in fault data and the need for robust predictive models that generalize well across different projects. In this research, we delve into these challenges and investigate the impact of class imbalance and model generalization on software fault prediction using cross-project analysis. Our study addresses three primary research questions: Firstly, we examine the critical issue of class imbalance in fault prediction, which poses a significant hurdle to accurate model performance. Through extensive experimentation with various classifiers on diverse datasets from different software projects, we highlight the variations in classifier performance and the necessity of addressing class imbalance for reliable predictions. Secondly, we evaluate the reliability of cross-project prediction, aiming to understand how effectively predictive models trained on one project can generalize to predict faults in other projects. We demonstrate the importance of training with datasets sharing similar characteristics with the target project for achieving reliable cross-project prediction. Thirdly, we analyze the impact of increasing training samples from different projects on prediction accuracy, emphasizing the benefits of utilizing cross-project analysis to enhance predictive model performance. In addition to addressing these research questions, we provide a comprehensive comparison of classifier performance metrics, including accuracy, precision, recall, and F1 Score. Our findings not only shed light on the challenges and opportunities in software fault prediction but also emphasize the importance of considering class imbalance and model generalization for developing robust and reliable fault prediction models. This research contributes to advancing the field by providing insights into effective modeling approaches and highlighting the motivation behind addressing these challenges.

Alleviating Class Imbalance Issue in Software Fault Prediction Using DBSCAN-Based Induced Graph Under-Sampling Method

Imbalanced Data Sets Classification Method Based on Over-Sampling Technique

Diversity based multi-cluster over sampling approach to alleviate the class imbalance problem in software defect prediction

Genetic algorithm-based oversampling approach to prune the class imbalance issue in software defect prediction

A New Improved Prediction of Software Defects Using Machine Learning-based Boosting Techniques with NASA Dataset

Support Vector based Oversampling Technique for Handling Class Imbalance in Software Defect Prediction

Tackling Class Imbalance Problem In Software Defect Prediction Through Cluster-Based Over-Sampling With Filtering

A Novel Imbalanced Data Classification Method Based on Weakly Supervised Learning for Fault Diagnosis

Imbalanced Big Data Oversampling: Taxonomy, Algorithms, Software, Guidelines and Future Directions

A Clustering-Based Resampling Technique with Cluster Structure Analysis for Software Defect Detection in Imbalanced Datasets

A Density-based Under-sampling Algorithm for Imbalance Classification

A Software Defect Prediction Method That Simultaneously Addresses Class Overlap and Noise Issues after Oversampling

A Novel Adaptive Undersampling Framework for Class-Imbalance Fault Detection

Performance evaluation of software defect prediction with NASA dataset using machine learning techniques

Adaptive Centre-Weighted Oversampling for Class Imbalance in Software Defect Prediction

An Empirical Study on the Effectiveness of Data Resampling Approaches for Cross-Project Software Defect Prediction

The Integrity of Machine Learning Algorithms against Software Defect Prediction

Software Fault Prediction Using Cross-Project Analysis: A Study on Class Imbalance and Model Generalization

A Novel Class-Imbalance Learning Approach for Both Within-Project and Cross-Project Defect Prediction.

A Survey of Different Approaches for the Class Imbalance Problem in Software Defect Prediction