Research On Cost Prediction Method Of Tested Module In Code Defect Detection

With the continuous development of software technology and scale,there are more and more problems in the process of software development.Code defect detection tools are favored by people because of their high efficiency of defect detection and highly automation.However,with the increasing size of code and the increasing complexity of code files,code defect detection tools are also faced with more challenges,such as hardware resources can not meet the requirements,detection time is too long and so on.At this time,the traditional single-machine code defect detection system can not solve these problems well,so it needs to use distributed scheduling method.If the approximate time and space cost of the tested module can be known before the distributed scheduling,it can be scheduled better,thus improving the efficiency of defect detection and the utilization of hardware resources.Based on the analysis of the process of code defect detection,this paper proposes a method to predict the cost of the tested module in code defect detection,and implements the cost prediction system of the tested module.According to the characteristics of the defect detection process,this paper extracts the time cost feature and space cost features,obtains the syntax tree sequence,time cost feature and space cost feature of the tested module through the cost prediction system of the tested module,and extracts the semantic features from the abstract syntax tree sequence through the deep memory network,The fusion feature is obtained by fusing the time cost feature and the semantic feature.Then the regression algorithm is used to predict the time cost of the fusion feature,and the regression algorithm is used to predict the space cost of the space cost feature.In this paper,10 open source c projects are evaluated,when using the cost prediction system of the tested module to predict the time cost,the experimental results show that the mean absolute error(MAE)is 0.0136 lower than that of the traditional regression model,and the mean square error(MSE)is 0.0085 lower than that of the traditional regression model;When using the cost prediction system of the tested module to predict the space cost,85.1%of the data whose error ratio is less than 20%.In general,the proposed method has a relatively good performance in cost prediction.