A Software Defect Prediction Method Based on Module Dependency Graph

Abstract

This disclosure proposes a software defect prediction method based on the module dependency graph, which identifies the defect information of the software module according to the version information of the software to be analyzed; establishes the software module dependency graph according to the dependency relationship between the software modules, and uses the developer as a Nodes in the module dependency graph; extracting internal features of the software module, extracting the dependency features of each node in the software module dependency graph by means of network representation learning, and forming a metric tuple with the internal features and the dependency features between the modules, Establish the historical defect library of the software according to the measurement tuple and the defect information of the module; use the historical defect library to train the defect prediction model for subsequent software defect prediction, and the defect prediction adopts the dynamic selection of classifiers based on local optimum The model automatically optimizes the parameters of the defect prediction model, and uses the result of the software module defect prediction model as the defect prediction result of the software to be analyzed. This method can improve the flexibility of constructing network node metrics and improve the effect of software defect prediction.

Description

technical field [0001] The invention belongs to the technical field of software quality assurance, in particular to a software defect prediction method based on a module dependency graph. Background technique [0002] Software defect prediction is a very important research topic in software engineering. Static software defect prediction technology based on metrics uses the historical data obtained from existing software modules to predict defects for new software modules to judge new software defects. Whether the module is defective or not, thus providing decision support for software projects. Most of the existing research on software defect prediction uses machine learning technology. Software defect prediction generally includes the following steps: 1) Mark module categories, software modules can be divided into two categories: defective modules and non-defective modules; 2) Extract module attributes, Use McCabe metric, McCabe metric, Halstead metric and other methods to...

AI Technical Summary

Problems solved by technology

[0006] Existing research mainly relies on user-defined structural feature metrics (such as degree statistics or centrality metrics) to describe the structural features of nodes, which lacks flexibility and makes it difficult to extract network node features.

Developers are also the cause of defects, and the existing research rarely takes this factor into account

Images