Research And Application Of Effort-aware Software Defect Prediction Based On Approximate Density

In modern society,various types of software products are inseparable from people's daily lives.With the continuous development of the Internet and communication technologies,the scale of software products and the complexity of their architectures are getting higher and higher.The probability of software defects and the harm caused by them are also paid more and more attention by researchers.In order to ensure the quality of software products,software defect prediction technology is applied to detect defects generated during the software development process.However,the traditional software defect prediction technology research ignores the test priority issue caused by the limitation of test resources and the difference in workload between modules.Therefore,effort-aware defect prediction methods are proposed and used to reduce the workload of testers and improve the utilization of test resources.For the context that the existing classification model does not perform well on the problem of effort-aware defect prediction,this paper proposes an effort-aware defect prediction method based on approximate defect density from the perspective of defect density.In heterogeneous effort-aware defect prediction scenarios,considering the difference in the feature distribution of the source and target project,this paper combines the approximate defect density prediction strategy with the kernel canonical correlation analysis method to propose a heterogeneous effort-aware approximate defect density prediction method ADPKCCA.In this paper,comparative experiments were conducted in isomorphic and heterogeneous scenarios to verify the effectiveness of the approximate defect density prediction strategy and ADPKCCA.Finally,an effort-aware defect prediction system was designed based on the approximate defect density prediction models in the above two prediction scenarios.The main work of this paper is as follows:(1)In this paper,the field background of software defect prediction and the research status of effort-aware software defect prediction are introduced and the research work of this paper is shown;(2)The approximate defect density index is proposed,and the approximate defect density prediction model is constructed.In isomorphic effort-aware defect prediction scenarios,comparative experiments were conducted with true defect density regression models and classification models to verify the effectiveness of approximate defect density indicators;(3)For the prediction of heterogeneous effort-aware software defect prediction,this paper proposes the ADPKCCA model.The heterogeneous defect prediction model currently performs unsatisfactoryly in effort-aware defect prediction problems,and even worse than simple unsupervised models.This paper combines the approximate defect density prediction strategy with the kernel canonical correlation analysis method to propose a heterogeneous effort-aware approximate defect density prediction method ADPKCCA.ADPKCCA is compared with existing supervised models and unsupervised models with good performance.Experimental results verify the effectiveness of ADPKCCA in heterogeneous scenarios;(4)Based on the effectiveness of the approximate defect density prediction strategy,an effort-aware EADP software defect prediction system is designed.This article introduces the requirements analysis,architecture and functional module design and prototype implementation of the defect prediction system.