Estimating Release Time and Predicting Bugs with Shannon Entropy Measure and Their Impact on Software Quality

Talat Parveen, H. D. Arora

Abstract

As large amount of software repositories are available, the quantification of code change process is made possible and software engineering process had been paced up over a period of time. These repositories which include code change process information, bugs, and details about developers are abundantly used by researchers to fetch information which are important for improving software quality. We presume that a complex code change procedure incompatibly affects software quality. Code change process affects the quality of software and hence the software cost is affected as well.  We developed a system in which data derived from the change history of each software release version is taken under consideration. This analysis shows that history change complexity metrics are prominent in predicting bugs in the software system in comparison to classical predictors of faults i.e., prior alterations, prior defects etc. Source code change data of software releases has been fetched from github repository for over a fifteen years of time, which includes 151 releases. History complexity metrics for the code change and bugs registered are used for predicting the release time and future bugs in the software release. To the data statistical multilinear regression model is utilized in predicting the software release time and estimated bugs of a software based on the history complexity metric in various releases. The performance of the model had been compared using performance $R^2$, $RMSE$ and $MAE$ values.