Predicting the complexity of code changes using entropy based measures

Abstract

Changes in software source codes are inevitable. The source codes of software are frequently changed to meet the user’s enormous requirements. These changes are occurring due to bug repairs (BR), enhancement/modification (EM) and the addition of new features (NF). The maintenance task becomes quite difficult if these changes are not properly recorded. The versions of these frequent changes are being maintained using the software configuration management repository. These continuous changes in the software source code make the code complex and negatively affect the quality of the product. In the literature, the complexity of the code changes has been quantified using entropy based measures (Hassan, in: Proceedings of the 31st international conference on software engineering, pp. 78–88, 2009). In this paper, we have proposed a model to predict the potential complexity of code changes using entropy based measures. The predicted potential complexity of code changes helps in determining the remaining code changes yet to be diffused in the software. The proposed model has been validated using seven components of web browser Mozilla. The model has been evaluated using goodness of fit criteria namely R squared, bias, mean squared error, variation, and root mean squared prediction error (RMSPE).The statistical significance of the proposed model has been tested using χ2 and Kolmogorov–Smirnov (K–S) test. The proposed model is found statistically significant based on the associated p value of the K–S test. Further, we conclude that the rate of complexity diffusion due to BR is found higher in four cases namely Bonsai, Mozbot, tables and XUL. The other components of Mozilla namely AUS, MXR and Tinderbox show increase in complexity due to EM and NF.