Software defect prediction method based on multi-granularity nodes

Abstract

The invention relates to the field of G06F11 / 36, in particular to a software defect prediction method based on a multi-granularity node, which comprises a code analysis step, a vector conversion step, a feature extraction step and a defect prediction step, and is characterized in that DL-Generated features are extracted from codes by adopting a DBN and a CNN respectively, a model is constructed, software defect prediction is carried out, and a defect prediction result is obtained. The selection of the node granularity is studied and discussed to obtain a proper granularity selection, so that the effectiveness of deep learning feature extraction can be further improved, the effectiveness is compared, and the proper node granularity is selected for software defect prediction. The method provides reference for software defect prediction researchers to select abstract syntax tree granularity, can effectively assist developers in judging and identifying potential risks and software defects in a new version software project in a software development process, and has very high practical application value.

Description

technical field [0001] The invention relates to the field of G06F11 / 36, in particular to a software defect prediction method based on multi-granularity nodes. Background technique [0002] As the scale of modern software continues to increase, the complexity of software increases, so the tasks of debugging and testing become more difficult in the process of software development and maintenance. In recent years, relevant researchers have begun to explore software defect prediction methods to assist and help developers find and predict software defects. Software defect prediction technology is based on machine learning methods, using historical defect data to build models to predict whether new codes are prone to defects. An excellent software defect prediction model can effectively count and predict the number and distribution of defects in the software system, so as to help the development and testing teams understand the software quality status, so as to rationally allocat...

AI Technical Summary

Problems solved by technology

However, such Handcrafted features usually cannot fully contain the semantic and structural information of the code, which may lead to poor prediction performance of the software defect prediction model. If the selected node granularity is not suitable for the software defect prediction task, the prediction effect of the prediction model may not be good. precise

Images