Software Change Prediction to Support Project Planning

Changes that occur in software are usually tracked and managed through version control systems. These version control systems contain data about changes in the form of change log files. Researchers have used this data to predict number of changes for future months for particular components of software. There is a need to carry out work on prediction of change size in addition to number of changes, because change size is a very influential factor in determining project size as it can cause an increase or decrease in it. In this dissertation, we focus on prediction of change size along with number of changes for particular components and developers. This research work includes experiments for prediction of average change size and number of changes for components and developers. We used three prediction techniques (ARIMA, Double Exponential Smoothing, and Polynomial Regression) for this purpose.