Proactive service selection based on acquaintance model and LS-SVM

Current service selection is unable to perform proactively. When a service provider overloads, the services list is ever-lengthening, which leads to backlog and failure of service composition. To compensate for this deficiency, this paper improves the proactive service selection. In this strategy, the service provider analyses a time series of services received to forecast the backlog and consign services to the others through a negotiation process. The least squares support vector learning is used to predict a random list of services, and an acquaintance model (AM) makes a consigner allocate backlog services to other providers with high degree of relationship. The backlog of services by forecasting is entrusted to the provider who can implement the same service, and negotiation between the providers with the same role would allow generation of a new service selection solution before a fault occurs. Experiments showed that the least squares support vector machine (LS-SVM) algorithm was more accurate in predicting a services list and a negotiation mechanism using AM decreased communication time effectively, which improved the success rate of service selection and reduced the execution time of service composition.