Reinforcement Learning Recommendation Algorithm Based on Label Value Distribution

Reinforcement learning is an important machine method and has become a hot popular research direction topic at present in recent years. The combination of reinforcement recommendation system, very application scenario application, always received close attention from researchers all sectors society. In this paper, we first propose feature engineering based on label distribution learning, which analyzes historical behavior analyzed constructs, whereby vectors are constructed for users products via learning. Then, algorithm value proposed. We designed the stochastic process process, described user’s state interaction (by including information their explicit implicit state), dynamically generated product recommendations through user feedback. Next, by studying hybrid strategies, combined dynamic static to fully utilize achieve high-quality algorithms. Finally, was validated, various relevant baseline models were compared demonstrate effectiveness study. With study, actually tested remarkable advantages design nonlinear expectations other homogeneous individual models. use systems with considerably increased accuracy, data utilization, robustness, model convergence speed, stability systems. incorporated idea into implementation main practical improved that its performance more accurate than same level computing power level. Moreover, due higher amount enhanced contains, it provides theoretical support basis can be used services, many prospects.