A deep learning approach to software evolution. (2018)

• Record Type: Journal Article
• Authors: Zheng, Shang; Yang, Hongji
• Is Part Of: International journal of computer applications technology. Volume 58:Number 3(2018)
• Page Start: 175
• Page End: 183
• Publication Date: 2018
• Subjects: software evolution -- deep learning -- machine learning -- topic model; Technology -- Data processing -- Periodicals; 620.00285
• Languages: English
• ISSNs: 0952-8091

Abstract

Software evolution techniques should be made as important as software development techniques. One possible way to help with the situation is to learn from software development, along with learning from software evolution techniques. The breakout of Machine Learning and Deep Learning (ML&DL) is becoming popular in technology and should be studied for being made available for servicing software evolution. Open source projects provide an open defect repository to which users and developers can report bugs. It is a challenge to document bug reports to the appropriate developers. In this paper, we apply deep learning approaches and a topic model to learn the features of defect reports and then make recommendations. Compared to the traditional machine learning approaches, the proposed approach based on deep learning can perform better in accuracy and assign defect reports to developers more effectively and correctly along with the dataset increasing.