Software Fault Prediction based on GSO-GA Optimization with Kernel based SVM Classification

The two main challenges of Software Defect Prediction are high dimensionality and class imbalance. The first challenge is high dimensionality where the number of features extracted from software modules becomes larger due to growth in size and complexity of modern software systems. The extracted features may be redundant or irrelevant. The problem of high dimensionality can be solved by an important pre-processing procedure that is feature selection. The second challenge is class-imbalanced data, where the major defects in a software system are found in few modules. The earlier methods failed to provide good solution for class imbalance and high dimensionality. Also, the prediction accuracy of the existing methods is low. To overcome these major drawbacks an optimal Group Search Optimization –Genetic Algorithm (GSO-GA) based fault prediction in software testing with kernel based SVM classification is used for improving the software quality. To evaluate the performance of the proposed approach that is hybrid Group Search Optimization-Genetic Algorithm (GSO-GA) based fault prediction is compared with existing Group Search Optimization (GSO). The GSO-GA based models have produced better results and accuracy in terms of existing software metrics like Average Percentage of Faults Detected (APFD), Problem Tracking Reports (PTR), and time and memory usage. From the experimental results, we observe that the average percentage of faults detected in our proposed approach is higher than the existing method.