A framework to assess the effectiveness of quality assessment model developed using class level metrics

Abstract

System and class level software metrics are often considered for predicting fault-prone modules in a software during the analysis and design phase of object-oriented software development life cycle (SDLC). However it is further observed that class level metrics also provide a good amount of insight on fault prediction. This study focuses on developing various fault prediction models based on public datasets. In order to validate efficiencies of prediction models for predicting fault proneness are often validated. To achieve this, a cost evaluation framework has been proposed to evaluate the effectiveness of the fault prediction models. This framework, is based on the classification of classes into faulty or not-faulty ones. From the obtained results, it is observed that fault prediction is useful for the projects with the percentage of faulty modules less than a certain threshold value. From the proposed models, it is also observed that no single model is sufficient to provide the best result (effective cost); but an attempt in this direction helps for an in-depth analysis.