Case difference heuristic adaptation method based on deep reinforcement learning

• Published in: Expert systems with applications Vol. 270; p. 126545
• Main Authors: Yan, Aijun, Cheng, Zijun
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 25.04.2025
• Subjects: Case adaptation; Case difference heuristic; Deep deterministic policy gradient; Deep Q-network; Deep reinforcement learning; Learning-evaluation-revision; Deep deterministic policy gradient; Deep reinforcement learning; Case adaptation; Deep Q-network; Case difference heuristic; Learning-evaluation-revision
• ISSN: 0957-4174
• DOI: 10.1016/j.eswa.2025.126545

To address the bottleneck problems of case adaptation knowledge acquisition and learning and the difficulty of simultaneously applying the network structure to multi-attribute case representation, this paper proposes applying deep reinforcement learning (DRL) to the learning of case difference heuristic (CDH) adaptation knowledge and implementing the generation process of a case adaptation solution based on the “learning-evaluation-revision” idea. The method first establishes the connection between DRL and the CDH adaptation method and then introduces the corresponding principles. Next, the CDH adaptation algorithms of deep Q networks (DQN) and deep deterministic policy gradient (DDPG) are given. The “evaluation-revision” process of adaptation is implemented according to the intelligent agent-environment mechanism of DRL. Finally, experimental verification is carried out on public datasets and actual solid waste data. The results show that the proposed method can effectively adjust case solutions to adapt to new problems, significantly improving the problem-solving quality of case reasoning and achieving good effects in actual applications.