GraphSAGE and LDA-based software defect prediction method

• Main Authors: CHEN XIANG, JU XIAOLIN, SHEN HAO, LU YURONG, SHEN YIHENG
• Format: Patent
• Language: Chinese; English
• Published: 26.05.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention provides a defect prediction method based on GraphSAGE and LDA, belongs to the technical field of computers, and solves the problem that a manual feature method does not consider natural language information software defect prediction of code items. According to the technical scheme, the method comprises the following steps: (1) collecting a defect instance data set from Githhub, and carrying out preprocessing operation on the data set; (2) extracting subject description information of the project by using LDA; (3) analyzing the code snippets into an abstract syntax tree, and extracting a relation matrix between tokens; (4) the token and the subject term information are coded into feature vectors through BERT, and the vectors are spliced; (5) inputting the vector into GraphSAGE training to obtain the feature representation of each node; and (6) inputting the node representation into an mlp classifier, and performing defect prediction. The method has the beneficial effect that the reliability and