A Study of Innovative Class Imbalance Dataset Software Defect Prediction Methods

Abstract

Data mining for software defect prediction is the best approach for detecting problematic modules. On-hand classification methods can speed up knowledge discovery on class balance datasets. Actual facts are not balanced since one class dominates the other. These are class imbalances or skewed data sources. As class imbalance increases, the fault prediction rate decreases. For class imbalance data streams, the suggested algorithms use unique oversampling and under-sampling strategies to remove noisy and weak examples from both the majority and minority. We test three techniques on class imbalance software defect datasets using four assessment measures. Results indicate that class-imbalanced software defect datasets can be solved.