Identification of latent variables using, factor analysis and multiple linear regression for software fault prediction

Abstract

Software practitioners develop models by considering the process of software fault prediction in the early stage of the software development life cycle in order to detect faulty modules. Various statistical and machine learning techniques are examined in the past for fault prediction. In this study, we have performed an empirical analysis of object-oriented (OO) metrics with a review of studies from the year 1996 to 2018 in the literature considering the statistical and machine learning techniques for software fault prediction. In this research, the concept of factor analysis and its sub-measures with regression are used to assess the capabilities for fault-proneness. The authors have also grouped the significant predictors i.e. CK and OO metrics using the factor analysis. In this paper, we have identified significant factor’s in five software’s datasets. The model is using the factor analysis with regression technique for estimating software fault-proneness. The results calculated prove the potential and capabilities of factor analysis for grouping important factors and using regression to identify the significant predictors. The experimental results obtained prove the ability of factor analysis with regression for predicting the susceptibility of software towards, the grouping of the components and effective use of the concept researchers and software practitioners. However, the significance and the application of the factor analysis with regression in software fault prediction are still limited and focus on these studies should be considered in order to generalize the results. On the basis of the results obtained, researchers are provided with future guidelines in this research work.