Deep Learning-Based Resolution Prediction of Software Enhancement Reports

Muhammad Ali Arshad, Zhiqiu Huang, Adnan Riaz, Yasir Hussain

Abstract: The automatic resolution prediction of newly submitted enhancement reports is an important task during the bug triage process. It can help developers automatically predict the resolution status of enhancement reports. The resolution prediction is still a manual process which is very time-consuming, costly, and laborious. To help software applications for the timely implementation of enhancement reports, we introduce a deep learning-based technique to predict the resolution of newly submitted enhancement reports automatically by using a summary and description of enhancement reports. We use Word2Vec and a deep-learning-based classifier that can learn the deep syntactical and semantical relationship between the words of enhancement reports. We use additional novel features from enhancement reports and customized tokenizer to save useful features. Experimental results show the proposed approach enhances the performance as compared to state-of-the-art approaches in resolution prediction and has an effective ability to predict the resolution status of enhancement reports.