BestRec: A Behavior Similarity Based Approach to Services Recommendation

Development efficiency is an important factor for the Internet-scale software produced through services composition. In this paper, we aim at improving development efficiency from two aspects. One is to improve services discovery efficiency through recommendation, and the other is to provide a mechanism to reuse an existing composite service. We propose BestRec, a behavior similarity based approach to services recommendation. With BestRec, the behavior captured by a composite service is described by a DFA-based service model. Based on the service model derived from the developing context of services composition and the existing composite services in the services repository, we calculate the behavioral similarity according to the relations between states of each existing service model and that of the developing service model. Through this approach we can automatically model the request of services discovery and recommend the most suitable composite services to developers in the light of their behavioral similarity. Finally, we illustrate the effectiveness of BestRec with a case study and experimental evaluation in the area of Web services testing applications.