Software visualization and deep transfer learning for effective software defect prediction

• DOI: 10.1145/3377811.3380389
• Publication Date: 2020-10-01T18:25:38Z
• Authors: Jinyin Chen; Keke Hu; Yue Yu; Zhuangzhi Chen; Qi Xuan; Yi Liu; Vladimir Filkov
• Keywords: Software bug; Feature (linguistics); Transfer of learning; Software visualization

Software defect prediction aims to automatically locate defective code modules better focus testing resources and human effort. Typically, software pipelines are comprised of two parts: the first extracts program features, like abstract syntax trees, by using external tools, second applies machine learning-based classification models those features in order predict modules. Since such approaches depend on specific feature extraction learning classifiers have be custom-tailored effectively build most accurate models.