Reinforcement learning recommendation algorithm based on environment information exploration

At present, most recommendation algorithms based on reinforcement learning focus on designing exploration strategies at the model level, however, they often neglect the full exploitation of environmental state information and the impact of action exploration on user experience. In order to effectively use environmental state information to guide action exploration, this paper proposes a reinforcement learning recommendation algorithm based on environmental information exploration. The algorithm makes full use of environmental information to improve the personalized effect of recommendation by constructing preferentially selected strategy and organizing state space. In addition, the algorithm guides the exploration direction of the model according to the environment state representation and user experience by comprehensively considering the reward signals inside and outside the environment, by fully mining environmental information, the algorithm can reasonably explore user preferences.