Predicting the Number of Reported Bugs in a Software Repository
Hadi Jahanshahi (Data Science Lab at Ryerson University), Mucahit Cevik (Data Science Lab at Ryerson University) and Ayse Basar (Data Science Lab at Ryerson University).
The bug growth pattern prediction is a complicated, unrelieved task, which needs considerable attention. Advance knowledge of the likely number of bugs discovered in the software system helps software developers in designating sufficient resources at a convenient time. They may also use the information to take vital steps to increase the quality of the system and in turn customer satisfaction. In this study, we examine eight different time series forecasting models, including Long Short Term Memory Neural Networks (LSTM), auto-regressive integrated moving average (ARIMA), and Random Forest Regressor. Further, we assess the impact of exogenous variables such as software release dates by incorporating those to the prediction models. We also analyze the quality of long-term prediction for each model based on five different metrics. The assessment is conducted on the large open-source software application, Mozilla. The dataset is originally mined from Bugzilla and contains the number of bugs for the project from Jan 2010 to Dec 2019. Our numerical results indicate that LSTM is effective when considering long-run predictions whereas Random Forest Regressor enriched by exogenous variables performs better for predicting the number of bugs in the short term. In addition, our analysis provides insights on evaluating the trends in a bug repository.