Hybrid depth defect prediction method based on code snippet analysis

The invention relates to a hybrid depth defect prediction method based on code snippet analysis, and belongs to the technical field of computer software defect prediction. According to the method, firstly, based on a program slicing method of defect library key points, an open source software code unit set containing defects is vectorized, and features are expressed as a vector form which can be processed by a deep learning model; then, based on a defect prediction method of hybrid deep learning, the classification and prediction capabilities of a hybrid deep model are improved, and a defect prediction classifier is obtained through training; and finally, defect prediction is performed on the open source software based on the trained defect prediction classifier, and target code snippets are outputted in a classified manner. According to the method, the pre-designed defect library key points are taken as the entry point of program slicing, the code snippets containing defect characteristics are extracted from