| Services selection is a crucial step to the sharing and reuse of services. With rapiddevelopment of Internet Technologies, the number of candidate services with same functionalproperties but with different non-functional properties is increasing devastatingly. How to selectan appropriate service effectively has become one of the hot topics in the research field ofservices computing. Current service selection approaches are mostly based on the quality ofservice (QoS) released from service providers, or modified QoS with trust evaluation on serviceproviders. Focus of these QoS-based approaches is the trustworthiness of QoS data and littleattention is paid to whether personal characteristics and properties of service requesters willaffect selection process or not. Collaborating Filtering (CF) techniques are employed in someservice selection approaches and selection is made within service requesters with samepreference, which successfully improve the accuracy and correctness of selection by decreasingthe number of service recommenders. However, these CF–based selection approaches fail toconsider the trustworthiness of recommenders and cannot resist malicious feedback fromrecommenders.This paper introduces user correlation to embody the impact of personal characteristics ofservice requesters on selection, computes creditability of recommendation, and employs AnalyticHierarchy Process to decide the weight of each factor in service reputation. With the integrationof trust evaluation methods and CF techniques, a trustworthy services selection model based oncollaborative filtering is presented.Simulation experiments demonstrate that model can not only improve the efficiency ofservice selection but also effectively avoid malicious attack from service recommenders.Meanwhile, this paper design and realization of online medical diagnosis system, throughconcrete examples to verify the advantages of this algorithm. |
