Entropy based Bug Prediction using Neural Network based regression

Bug Prediction is an important research area in the field of software engineering. Researchers have developed and implemented a number of bug prediction approaches like past bugs, code churn, refactoring, file size and number of authors, etc and measured their performance. Various mathematical models have also been proposed by researchers for monitoring the bug detection and correction process. The bugs are introduced in the software mainly because of the continuous changes that occur in the software code. These continuous changes tend to make the code complex. The complexity of code changes, quantified by Entropy is used to predict bugs. In previous research, Statistical Linear Regression is used to construct bug prediction model. In this paper a Neural Network model of entropy based bug prediction is developed and compared with SLR model. It is observed that Neural Network based Regression (NNR) performs either better than or nearly equal to SLR.