Research On Complex Multi-attribute Evaluation Of Software Quality Based On Error Propagation Theory

Software quality refers to that the software meets the requirements of users.It is an important index to evaluate the quality of software products.With the rapid development of information technology,in recent years,all kinds of software came into being,so that rich software products gradually play an increasingly important role in people's production and life.The practicability and scientificity of software selection have been concerned for a long time.Therefore,software quality evaluation is an effective means to ensure the quality of software,but also an indispensable part of the software development process.At present,there are many problems in the complex multi-attribute evaluation of software quality in China,such as the mixed multi-attribute evaluation of multi kinds of uncertain evaluation information,including the unreasonable weight distribution in the evaluation process,and the information loss in information integration.Using scientific and effective evaluation methods to reasonably grasp and ensure the software quality has become the key to the problem,and it also has become a problem to be solved at this stage.This thesis focuses on the multi-attribute evaluation of software quality,as stated as followings.Firstly,this thesis presents the normalization methods of fuzzy language evaluation information,ranking information and interval number information,which can effectively solve the comparability problem of uncertain evaluation information.Interval number is used as the benchmark to represent all kinds of evaluation information,which effectively retains the fuzziness of language evaluation in the evaluation process and avoids information distortion in the process of information aggregation of different types of preference information,thus improving the authenticity of the evaluation.Secondly,in order to aggregate the evaluation information matrix given by experts based on error propagation theory,K-means clustering algorithm is introduced,and the weighting method of K-means clustering center matrix is used to solve the weights of experts.At the same time,based on the collected evaluation information matrix,the entropy weight method and TOPSIS are combined to determine the attribute weights,considering the non utility influence among attributes,and make the attribute weights more accurate.Finally,applying a gray correlation degree method,utilizing the information entropy in attribute information,by calculating and measuring the distance of each software correlation degree zone to the ideal correlation degree interval number,the integrated quality of the software is determined.At the same time,the illustration example shows the feasibility and practicability of the proposed approach to the software quality evaluation proposed in this thesis.