Research On Software Defect Evaluation Model Based On Complex Networks

As the application area of the software continues to expand,the demand for software service is increasingly high,the scale and complexity of the software are increasing,and in software systems,the existence of various kinds of software defects makes it difficult to guarantee software quality.The traditional software metric methods focus on local features,and the size of the metric is small,which leads to the limitation of the software defect evaluation model based on the traditional software metric methods.The discovery of complex networks in the software system structure provides a new idea for the research of software structure feature metrics.From the point of view of software network,software structure features can be measured by using the theory and methods of complex networks to find complex or important software entities.Therefore,the research of software defect evaluation model based on the theory and methods of complex networks has important significance.From the point of view of software network,this paper uses the theory and methods of complex networks,combined with structured program metric methods and object-oriented software metric methods,to measure software structure features.This paper regards the software defect evaluation as three classification or regression sub-problems,including whether there are defects in each class,software defect severity evaluation and software defect priority evaluation.This paper constructs the software defect evaluation model based on complex networks through machine learning based on s software structure feature metrics and software defect data.The model can evaluate Software defects on the class hierarchy,and predict whether there are defects in each class,software defect severity and software defect priority.In Windows 7,this paper constructs the software defect evaluation model based on complex networks for open source software Hibernate.After comparison of experimental data,this paper selects Random Forest model as each sub-model.The experimental results show the validity and accuracy of the software defect evaluation model based on complex networks.