Machine learning based approval prediction for enhancement reports

In modern times, the maintenance of the software application plays a vital role in its success. Software applications obtain enhancement requests on a large scale to fulfil user requirements through different Issue Tracking Systems. Issue tracking system provides an effective way for keeping the bugs records in the software development system. Conventionally, developers used to manually check these requests themselves. However, manual inspection of these requests turns out to be a boring, hectic and time-consuming activity.Therefore, there is dire need of developing an automatically prediction system, that can help in decision making for further improvement. In this work, we propose a Support Vector Machine-based classifier to automatically approve or reject an enhancement report. Our approach can be divided into different steps. Firstly, we perform the pre-processing on each enhancement report using natural language processing (NLTK) techniques. Secondly, we generate a feature vector for each pre-processed enhancement report. Finally, we train a Support Vector Machine-based classifier that automatically predicts the rejection or approval of the enhancement report. In order to have a thorough analysis, we also evaluate and compare other well-known machine learning algorithms e.g. Multinomial Naive Bayes and Logistic Regression. We use a well-known open-source dataset extracted from the Bugzilla software application for our experiments. Our experiments suggest that Support Vector Machine-based classifier outperforms other approaches and achieves high accuracy on 35 different open-source applications which include 40,000 enhancement reports. The evaluated results of tenfold cross-validation show that the proposed approach can increase the accuracy as compared to the state-of-the-art accuracy. We believe that our approach will help developers save time and address user-requirements in a more efficient manner.