Service Recommendation Using Customer Similarity and Service Usage Pattern

Abstract

With the increased number of web services advertised on the internet, it is becoming vital to resolve typical problems of service recommendation. Although service recommendation has been studied by researchers in recent years, existing methods have remarkable achievements on offering single service recommendation, not only considering functional features of web services but also non-functional features. However, the customers usually adopted composite services to satisfy complex and coarse-grained requirements. The traditional service recommendation does not have much concern about composite services. Through a long period of usage, the dependencies among composite services are hidden in historical usage records. In reality, these dependencies have great influence on the quality of service recommendation. To improve the effectiveness of service recommendation, this paper proposes a novel service recommendation approach based on service usage patterns. Firstly, the similar customer group of target customer is identified through the personal attribute based clustering and similarity of rating preference, Secondly, service usage patterns of the similar customer group are mined based on the variant of Generalized Sequential Patterns (GSP) algorithm, Thirdly, promising services are recommended for the target customer according to the matching degree between previously used services and service usage patterns, Finally, experimental results verify the efficiency and effectiveness of our approach.