Missing QoS-values predictions using neural networks for cloud computing environments

Cloud computing environment is influenced by user-dependent quality of service (QoS) parameters in evaluating the performance of Web services apart from others factors. Among the performance QoS parameters, mainly response-time and throughput could be modulated to provide very efficient services for cloud users. As per user´s requirement, the service provider´s recommendation of appropriate Web services to the end-users with proper QoS satisfaction is one of the critical issues. This can be recommended to end-users in the Service Level Agreement (SLA) under Web Service Modeling Ontology (WSMO) of WS-Policy. Generally, the matrix of collected QoS parameter values is sparse and the accurate prediction of the missing QoS values is important for the recommendation of appropriate web services to the end users. To address this issue, we worked out an artificial neural network model for the prediction of missing QoS-values using past QoS performance parameter data. In this current work, the performances of different learning algorithms of ANN are analyzed for enhanced prediction of QoS performance values. The ANN model with Bayesian-Regularization is found to be better performing when compared to other learning algorithms.