Genetic Evolution-Based Feature Selection for Software Defect Prediction Using SVMs

doi:10.1142/s0218126622501614

Genetic Evolution-Based Feature Selection for Software Defect Prediction Using SVMs

2022-06-23

0

PRE

AI

DOI：10.1142/s0218126622501614

原文链接

原文求助

摘要

En 中文

Software Defect Prediction (SDP) involves the early detection of fault-prone modules and reduces the testing efforts and cost. Support Vector Machine (SVM)-based SDP classifiers use large amount of high-dimensional data, and hence feature selection (FS) is applied for better accuracy. Search-based feature selection is found effective to improve the efficiency of predictors. This paper proposes a genetic evolution (GeEv) technique to select features. The GeEv technique involves introduction of diversity at intermediate level by the genetic evolution of random offspring with better survival capability. GeEv searches the feature space for an optimal feature subset using the performance of classification and number of features selected as the fitness function. The FS is modeled as an optimization problem and optimal solution is sought using GeEv. The performance is compared with baseline technique. From the experimental results, it is shown GeEv outperforms the competing FS approach and can achieve better accuracy than others statistically.

Genetic Evolution-Based Feature Selection for Software Defect Prediction Using SVMs

摘要

期刊

机构

引用论文