Research On Service Reliability Prediction Based On Deep Belief Networks

With the continuous development of Internet and related network technology,more and more enterprises choose to release their products in the form of service to the network for users.However,due to the increasingly complex needs of users,single function services can not completely solve the problem of users.SO A based service system combines multiple services to meet the complex needs of users.However,because the network environment is always dynamic,it poses a great threat to the stable operation of the service,and the change of individual service will bring cascading effect and then affect the whole system.Therefore,the study of service reliability prediction is of great practical significance.Reliability prediction is mainly aimed at predicting the service reliability of the service in a certain period of time to ensure the stable operation of the service system.However,because of the uncertainty of the network environment,the parameters used to predict the reliability of service are difficult to obtain.These factors lead to greater difficulties in the prediction of service reliability.In this paper,we use time series to partition the running parameters and reliability of limited services,and discover the inner correlation by using the deep belief network.The Restricted Boltzmann Machine,as a two layer neural network,has good performance in feature extraction and classification.The Deep Belief Network can be defined as a stack of multiple Restricted Boltzmann Machines which can play a better performance.As a single hidden layer feedforward neural network,a single hidden layer feedforward neural network randomly selects the hidden layer output according to the probability distribution,and avoids the iterative tuning of the parameters,so it has a faster learning speed.(1)We use Deep Belief Network model to predict the reliability of Web services in the future time,and analyze the effectiveness of the method through a series of comparative experiments.(2)The Extreme Learning Machine is introduced to solve the problem of service reliability prediction by combining it with the traditional deep belief network.Because the structure of the Extreme Learning Machine is simple and easy to learn,it can improve the efficiency of the algorithm faster than the traditional Deep Belief Network,and we have verified it through experiments...