Composite Service Recommendation Based on Bayes Theorem

The number of web services increased exponentially in the past decade. Identifying a set of best candidates from vast services is the first step of composite service recommendation. Much of the previous work focused on the accuracy and efficiency in service matchmaking and optimization. Little attention was paid to the logs of web service execution, which contain diverse information such as service functionality, QoS record, execution order, etc. In this paper, we present how we apply Bayesian approach in analyzing service execution logs and then recommend an optimized service sequence based on the original service sequence. Compared with the existing methods, the approach has two main advantages: first, the resulting service sequences are more robust and compatible; second, the result has higher (both explicit and implicit) quality. The authors also propose three algorithms that are based on the recommendation approach. At the end, the authors show the experiment results to illustrate how this work helps facilitate composite service recommendation.