Classifiers Recommendation System for Overlapped Software Defect Prediction Using Multi-Label Framework

Shivani Gupta and Kusum Lata Jain

The recent research issue is the quality of software defect prediction (SDP) data set which is available publicly. The major quality issues in software defect prediction data set are class imbalance and class overlap. The class overlap increases the difficulty for the classifiers to learn and predict the defectiveness of source files accurately. In this paper our primary goal is to recommend the classifier for overlapped Software Defect Data sets (SDP). The recommendation of classifiers is based Multi label framework on which we use the meta learning extracted from defect data sets. The results obtained shows that the our proposed framework provides the recommendation of classifiers suitable for new instances in SDP data set that are highly overlapped.

Volume 12 | 03-Special Issue

Pages: 1472-1478

DOI: 10.5373/JARDCS/V12SP3/20201399