Adaptive Web Services Composition Using Q-Learning in Cloud

Plenty of web services are emerging in clouds. They are distributed, heterogeneous, autonomous and dynamic. These characteristics may make a composite service unstable and inflexible. To adapt to this environment, we propose a machine learning strategy that is developed for and applied to web service composition. This way, the composition framework continually learns which web service candidates are currently best suited to be selected and composed to fulfill more complex tasks. Since the learning process is not stopped, the framework is able to adapt its composition strategies to changing conditions in dynamic environments. A case study is given and the learning algorithm is evaluated and compared to the results of related work, which shows that our method improves the success rate of service composition.