Filter-Wrapper based Feature Ranking Technique for Dynamic Software Quality Attributes

This article presents a filter-wrapper based feature ranking technique that is able to learn and rank quality attributes according to new cases of software quality assessment data. The proposed feature ranking technique consists of a scoring method named Most Priority of Feature (MPF) and a learning algorithm to learn the software quality attribute weights. The existing ranking techniques do not address the issue of redundancy in ranking the software quality attributes. Our proposed technique resolves the redundancy issue by using classifiers to choose attributes that shows high classification accuracy. Experimental result indicates that our technique outperforms other similar technique and correlates better with human experts.