Trustworthiness Prediction For Web Services Based On Selective Neural Network Ensemble Method

With the popularization of service computing technology,Web services are widely used as an important software resource on the Internet.In the practical scenario,the trustworthiness of Web services becomes an important goal that people need to consider when choosing and recommending Web services.To this end,it is important to evaluate and predict the trustworthiness of Web service effectively in the application process.In general,quality of service(Qo S)is an intuitive and important embodiment of the trustworthiness of Web services.Therefore,it is necessary to make a comprehensive analysis of QoS and to predict the trustworthiness of service.This paper attempts to solve the problem of Web service trustworthiness from the perspective of machine learning.In our solution,the selective neural network ensemble method is adopted to predict the trustworthiness of Web service.The main idea is to combine BP neural network,selective ensemble learning and particle swarm optimization together to obtain the better prediction performance.Firstly,the partial QoS records with trust rate are used to train BP neural networks to generate multiple candidates.Then,the particle swarm optimization is used to search for the weights of ensemble learning.The selective ensemble of neural networks is realized according to the optimal weight solution.In the ascept of technology implementation,two kinds of ensemble patterns,namely PSO-SEN algorithm and QPSO-SEN algorithm,are proposed based on the different coding of ensemble weights.The feasibility and effectiveness of the trustworthiness prediction technology for Web service based on selective neural network ensemble and the influence of parameters on the algorithm are verified by the experimental analysis of the public data set.Compared with other typical methods,the experimental results show that our solution has obvious advantages in prediction accuracy.It has low sensitivity to classifier integration,population size and classifier hidden layer nodes,which proves good robustness of the algorithm.The trustworthiness prediction of Web services is helpful to provide the guidance of trustworthiness characteristics for users when making service selection for a number of Web services with the same function,so that can make a scientific and rational service selection decision.In addition,the research of hybrid forecasting algorithm based on neural network,ensemble learning and intelligent searching can provide some reference for similar decision-making problems.