Reinforcement learning belongs to machine learning, with the autonomous learning method that can improve its action policy by interacting with environment. In order to improve the efficiency of grid service selection, a new approach based on improved reinforcement Q-learning for dynamic grid service selection is proposed. The environment of Grid service selection is a nondeterministic Markov decision processes (MDPs), and the study of grid service selection learning method is a challenge to current reinforcement learning which is based on MDPs. This paper proposes a correlative improved method for dynamic grid service selection. The experiment results show that the novel method is more effective in some aspects than traditional ones. Therefore it provides a good solution to select grid service.
展开▼