CROSS PROJECT LEARNING FOR FAULT PREDICTION WITH IMBALANCED DATA

Abstract

This study delves into the challenges of dealing with contradictory evidence and generalizing models. In addition, it investigates how incorporating other research efforts could enhance software failure prediction. The inability of traditional failure prediction methods to be highly task-specific stems from the fact that not all tasks share the same data. Machine learning procedures, data resampling methods, and feature selection tactics can help you overcome these challenges and make more accurate predictions. This project has two main goals: first, to improve model training and second, to investigate the usage of multiple datasets in order to hasten problem identification and ensure that solutions operate with varying software configurations. The findings have the potential to enhance and contextualize software quality assurance methods.