A Method for Predicting Perishing Services in a Service Ecosystem

With the wide adoption of Service-oriented Architecture, we can observe a rapid increase of web services these days. These services with their compositions produced by consumers form a service ecosystem. However as time goes by, some of the services are no longer available or says perishing due to the competition among the ecosystem. Obviously, compositions invoking these perishing services are also becoming unavailable. Thus, choosing the services which are more stable can help the consumers to product more valuable composition. The goal of this paper is to find a way to separate the potential perishing services in the ecosystem so that we can give suggestions to service consumers and help them develop more durable and valuable compositions. Firstly, we study the competition relation between services by their user tags and establish a service-service competition network. Based on the network analysis, we extract the common feature of perishing services and formalize this feature as percentage ranking (PR) of services. Finally, we propose a classification algorithm to predict potential perishing services. With the good performance in recall and precision rate, our algorithm is credible for identifying potential perishing services thus we can suggest the consumers to select more durable services for their compositions.