Research On Software Defect Technology Based On Date Measurement And Multi-objective Decisions

Software defects are unavoidable "by-products" in the process of software development.They not only affect the quality of software,but also lead to overexpenditure of development cost,out-of-control development schedule and even disastrous results.The software defects can not be discovered and eliminated simply by means of detection and verification.Therefore,software defect detection technology is indispensable.On the one hand,in the previous research,the focus of the software defect detection model was concentrated on the design of classifier,while the data complexity was ignored.On the other hand,the evaluation indexes of the detection model are diverse,so it is difficult to directly judge which model indexes are qualified or favorable.Therefore,it is very important to have stable evaluation technology.Aiming at the problems,this paper proposes a software defect detection method based on data complexity measurement and an evaluation method based on multi-objective decision:.(1)A software defect detection method based on data complexity measurement is proposed,the complexity metrics data in the data mining technology is firstly introduced and applied to the software engineering software defect detection model,The complexity metrics include 11 indexes are calculated on the NASA datasets.And then Three current relatively popular data mining algorithms of nearest neighbor classification algorithm,the decision tree algorithm and naive bayes classification algorithm are used all the data sets,to get multiple evaluation indexes.Finally we discusses the relationship between complexity of the data set and software defects detection results.The results show that the data complexity measurement can provide effective information for classifier selection and provide a strong support for establishing a stable detection model.(2)Based on multi-objective decision algorithm,amethod named Fuzzy AHP is proposed to evaluate software defect models.Analytic Hierarchy Process(AHP)and fuzzy mathematics are the main methods used in the multi-objective decision algorithm,which can solve the problem of multi-input and multi-output.In the later experiment,AHP is mainly used as the modeling object,as well as Fuzzy AHP which involves some deformation methods of AHP.Firstly,the qualitative and quantitative analysis of the software defect models conducted by using the analytic hierarchy process(AHP).And then the multi-decision evaluation method is established by using several measures in machine learning and the obtained data is used to establish the optimized evaluation algorithm.Finally,The multi-objective attribute decision model,Fuzzy AHP model was applied to perform on the evaluation metrics.The experimental results show that evaluation metrics finally obtained by detection models can be transformed into quantitative analysis through Fuzzy AHP method,and the evaluation of the solution can be found among the five criteria to determine whether the solution meets the expectation.