利用优先级双重深度强化学习的自适应Web服务组合方法

An Adaptive Web Service Composition Method Based on Priority Dual Depth Reinforcement Learning and POMDP

摘要:

针对大规模Web服务组合在动态环境下难以实现高可靠性、高动态适应能力的问题，提出一种结合优先级双重强化学习和POMDP的自适应Web服务组合方法。首先，采用POMDP对大规模Web服务组合优化策略进行建模，简化了组合优化分析的步骤，提高了大规模Web组合服务的效率；然后，在POMDP基础上，利用双重深度强化学习方法对优化策略进行分层重构，并求取最优解，提高了组合服务对动态服务环境的适应能力；实验结果表明，与现有优秀方法相比，所提方法在可靠性、效率和动态环境适应能力方面均有显著提升。

Abstract:

In view of the problem that large-scale Web service composition is difficult to achieve high reliability and high dynamic adaptability in dynamic environment, an adaptive Web service composition method combining priority dual reinforcement learning and POMDP is proposed. Firstly, POMDP is used to model the large-scale Web service composition optimization strategy, which simplifies the steps of composition optimization analysis and improves the efficiency of large-scale Web composite service. Then, on the basis of POMDP, by using dual depth reinforcement learning method, the optimization strategy is restructured and the optimal solution is obtained, which improves the adaptability of composite service to dynamic service environment. The experimental results show that compared with the existing excellent methods, the proposed method has obvious improvement in reliability, efficiency and dynamic environment adaptability.