Research On Software Defect Prediction Technology Based On Deep Learning

Due to the complexity of the software,software defects have increased,the disaster which is caused by these software defects often appears.So,in order to reduce the software disaster,software defect prediction technology has become a research hotspot in the field of computer science and technology.In the process of building a software defect prediction model,feature selection algorithm has great effect on the accuracy of software defect prediction model.However,the traditional feature selection algorithms such as PCA and LDA have a lot of limitations in deeply learning data feature and resisting the interference of noise or missing value.To solve these problems,in this thesis,two kinds of deep learning algorithms are proposed to construct software defect prediction model.At the same time,this paper proposes a complete framework of software defect prediction model,and developes software defect prediction system based on this framework.The main research work is as follows:Firstly,to solve the problem that the traditional feature selection methods,such as PCA and LDA,are unable to get the nonlinear relationship between characteristics,This thesis uses deep belief networks and SVM to construct software defect prediction model(DBN-SVM).Comparing with models build by using PCA and LDA,DBN-SVM software defect prediction model has higher prediction precision.Secondly,because deep belief networks cannot eliminate the noise and missing value which effect the accuracy of software defect prediction model.This thesis uses denoising autoencoders and SVM to built software defect prediction model(DA-SVM),compared with the DBN-SVM,DA-SVM software defect prediction model not only improves the prediction precision,but also enhance the robustness of the model.Thirdly,because data preprocessing and learning algorithm can also effect precision of the defect prediction model,To solve this problem,This thesis proposes a software defect prediction model framework which includes data preprocessing,feature selection and learning algorithm.At same time this paper developes the software defect prediction system based on this framework.The software defect prediction system can according to data sets select the best learning rule to build software defect prediction model and uses the model to predict software defect.