Design And Implementation Of Software Defect Analysis System Based On Text Mining

Software defect analysis is one of the important technical means to evaluate and improve software products and software processes,and has become a hot research topic in academia and industry.Aiming at the improvement of software process,this paper studies the method of text mining and statistical analysis to analyze the software defects.Through the text retrieval of software defect description,text clustering and text summary,the distribution,aggregation and repetition of software defects are excavated and analyzed.However,software defect description generally has the problems of redundant information and nonstandard syntax,and the distribution of noise words,synonyms and similar words is also very extensive.At the same time,the defect description for application software involves many kinds of objects,and the relationship between objects is also more complex.These make the feature selection,text abstract and text matching of software defect description challenging.In order to solve these challenges,this paper proposes software defect text mining technologies and methods,and carries on the system design,implementation and testing.The research work include:(1)This paper analyzes the application scene of software defect analysisfor software process improvement,and identifies the existing problems from its core activities.And then an overall approach of software defect analysis is put forward using text analysis and mining technologies.(2)According to the characteristics of software defect description,such as low quality,poor grammar,complex object relationship and dense synonyms and similar words,combined with deep semantic analysis,mutual information and document frequency statistics,this paper puts forward a feature selection and summary scheme for software defect description.On the basis of the preliminary summary,the deep semantic analysis based on dependency structure tree is proposed for feature selection and text summary.(3)A defect matching method based on feature and semantic mapping is proposed.According to the statistics of mutual information and document frequency,the text vectors needed for text retrieval are selected,and the attribute words closely related to the topic are selected as the text vectors for similar defect search and defect subdivision clustering according to the dependency structure tree.At the same time,the corresponding matching methods are designed for different text mining tasks,and the ontology information and synonymous information of words and word associations are mapped to the matching process.(4)It analyzes the functional and non-functional requirements of software defect analysis system,designs the overall architecture of the system with multi-layer architecture style,and implements the core modules of the system.The processing flows and class diagrams are described in detail.(5)The comprehensive function test and the performance test for the software defect analysis system are carried on the network management software defect data set.The test results show that the recall rate and precision rate of defect retrieval and similar defect search achieve 84.5% and 88.4%,81.9% and 91.4%,respectively.The precisionrate of defect subdivision clustering is 88.5%,and the precision rate of defect description extraction is 93.1%.The system meets the predetermined requirements.