Cross project defect prediction using hybrid search based algorithms

Abstract

Prediction of faulty modules in the software development cycle earlier helps to reduce the cost of software development. Test engineers give more attention to the faulty modules to remove any latent defect in the software. Most of the studies available in literature have used historical data related to the same projects for identification of faulty modules; however availability of historical data for new software projects is not possible. In case of new software projects, data for defect prediction is taken from similar types of projects developed earlier and this technique of defect prediction is called cross project defect prediction. In this study applicability of hybrid search based algorithms for cross project defect prediction is investigated. Performance of hybrid search based algorithms is compared for with-in and cross project defect prediction. Hybrid search based algorithms combine the advantages of search based algorithms with machine learning techniques. Results showed that hybrid search based algorithms are more suitable in case of cross project defect prediction in comparison to with-in project defect prediction.