Self-Adaptive Web Service Composition Based on Stochastic Context-Free Grammar.

We propose to model the adaptation of the service composition using the Stochastic Context-free Grammar (SCFG) in the loosely coupled and interoperable Service-Oriented Architecture. According to our approach, every concrete composition is regarded as an instantiation of the corresponding Stochastic Finite Automaton, which is regarded as the abstract composition. Analyzing the historical compositions in the users' logs, we acquire the probabilities of productions in the SCFG. And probabilistic functions are defined based on the structures of service composition. By means of incorporating the correlation degree among the services, the proper fault-handling strategy is determined using the adaptive decision algorithms. Furthermore, we demonstrate one self-adaptive framework monitoring the service composition execution. At last, the result of a case study shows our approach effectiveness.