A Hierarchical Diagnosis Method for Software Errors Based on Bayesian Network Reasoning

Abstract

The invention relates to a software error hierarchical diagnosis method based on Bayes network reasoning. The method comprises steps as follows: creating a function layer, applying MIC theory, improving the accuracy of statistical dependence, improving and applying a Laplace smoothing strategy, and solving test data rarefaction. Test prediction is performed, and whether a test case is a failure test case of the whole program is judged. Error location is performed, the doubtful degree of functions is calculated and sequenced, a statement level BNPDG of the corresponding function is created, the functions are inspected one by one according to the doubtful degree sequence, test prediction is performed on the statement level BNPDG of the functions, a bug function is found, error location is performed on the statement level BNPDG of the bug function, and the doubtful degree sequence of all statements in the function is obtained. The statements are checked one by one according to the doubtful degree sequence until the bug statement is found. Space consumption and time consumption of error diagnosis are reduced, and the problem of presence of local doubtful degree of calculation of RankCP and other technologies is solved.

Description

technical field [0001] The invention belongs to the technical field of software engineering software testing, in particular to a software error hierarchical diagnosis method based on Bayesian network reasoning. Background technique [0002] (1) Software testing technology [0003] A complete software development cycle is divided into six stages: problem definition and planning, requirements analysis, software design, program coding, software testing and operation and maintenance. Among them, software testing, as the last step before the software faces users, determines the quality of the software, and is particularly important in the entire software development cycle. With the continuous development of commercial software, the complexity of software is getting higher and higher, the development cycle is getting shorter and shorter, and users' requirements for software are becoming more and more stringent, which undoubtedly further enhances the importance of software testing...

AI Technical Summary

Problems solved by technology

[0019] Aiming at the problem that the existing error diagnosis method based on the program dependency graph can only diagnose local anomalies in the process of program execution, and needs to build a large-scale graph structure, the present invention designs a layered error diagnosis framework , first perform error diagnosis at the function layer to help program testers locate the specific function where the bug is located

Images