Research And Application Of File-Level Effort-Aware Software Defect Prediction

With the continuous development of software technology,software systems are becoming more and more closely linked to human production and life,and the scale of software systems is bigger and bigger.Software defect prediction system can help testers find defects in software system quickly.In the early stage of software development,finding and fixing defects in software systems will reduce the development cost of the entire software project.Nowadays,the scale of software system is getting bigger and bigger,and it is a very heavy work to check the source code file that may contain the defect.Moreover,for some software defect prediction systems with low prediction accuracy,the tester may check a lot of source code files and do not find real defects.This will combat the enthusiasm of the testers and reduce the efficiency of the work.Unlike traditional software defect prediction methods,effort-aware software defect prediction methods only need to check a small number of source code files to find most of the defects in the system.In order to minimize the effort of the tester,this paper proposes an effort-aware defect ranking algorithm for file level software defect prediction.Aim at reducing the precision of traditional defect prediction model after introducing the concept of effort-aware,this paper proposes a supervised defect prediction algorithm based on "ManualUP" and logical regression algorithm,which is based on the combination of "ManualUP" and logical regression.The traditional defect prediction model is less accurate under the evaluation index of effort-aware.Combined with "ManualUP" and logistic regression algorithm,this paper proposes a supervised defect prediction algorithm based on "ManualUP" and logistic regression.In the software defect prediction of effort-aware,the prediction accuracy of software defect prediction model is improved,and the validity of the algorithm is verified by experiments.Finally,a software defect prediction system is designed based on this algorithm.The main work of this article is as follows:(1)This paper introduces the background knowledge,research significance and research status related to the content of this paper,and puts forward the solution to the problem of low precision under the condition of effort-aware,and describes the research purpose and related technology of this paper.(2)In the defect prediction of file level,the software defect sorting algorithm of effort-aware is proposed.The algorithm builds a simple unsupervised model by using the software feature metric element,and only needs to check a few source code files to find most of the defects in the software system and reduce the effort of the tester.(3)A supervised defect prediction algorithm based on "ManualUP" and logistic regression is proposed.In view of the problem that the traditional software defect prediction algorithm is not accurate in the prediction of effort-aware,"ManualUP" is combined with the logic regression algorithm to improve the prediction accuracy and reduce the effort of the testers.(4)A supervised defect prediction algorithm based on ManualUP and logistic regression is designed,and a effort-aware software defect prediction prototype system is designed.The system requirements are introduced,and the prototype system is designed and implemented.