Reliability Prediction Of Cross-project Service System Based On Transfer Learning

With the development of Web services technology,more and more enterprises will publish their services to the network for users to use.Reliability prediction of service systems is one of the active research fields in software engineering.With the gradual improvement of people's demand for the reliability of service system,the traditional reliability prediction model for single service system can't meet the actual demand well.Therefore,the research of cross-project service system reliability prediction has strong theoretical significance and application value.This thesis researched a cross-project service system reliability prediction technology based on transfer learning to support and implement cross-domain transfer of the service system reliability prediction framework.Firstly,the reliability training is carried out within the source project based on the Bi-LSTM neural network.Then,the distribution difference between the source project and the target project is reduced by combining the method based on the Multi-kernel Maximum Mean Discrepancy,and the reliability prediction model of the cross-project is obtained.Service system reliability prediction tasks for cross-projects.The reliability prediction task for the cross-project service system is finally completed.In order to verify the validity of the model,this paper compared the influence of different factors on the reliability prediction results based on multiple experiments,and compared it with other common benchmark methods.Through multiple sets of comparative experiments,it is proved that the reliability prediction model BLPT-MMD proposed in this paper has significant advantages in improving the reliability prediction results of cross-project service systems.