Educational Psychological Behavior Analysis Based on Neural Hawkes Process and Reinforcement Learning

To solve the issues of complicated dynamic development of student psychology behavior in the scenario of online learning, inadequate capacity of modeling the temporal behavior dependencies, lack of precision of psychological state awareness, and low flexibility of personalized course recommendation, this study proposes an educational psychological behavior analysis and personalized recommendation framework that combines Neural Hawkes Process and Reinforcement Learning. This study relies on the modeling advantages of Neural Hawkes Process for continuous-time discrete events. It relaxes the non-negative constraint and continuous-time temporal encoding mechanism, and builds an encoder-dual-decoder hierarchical architecture. On this basis, this study combines a four-tuple course pre-filtering strategy to optimize the action space of Reinforcement Learning. The approach adopts a piecewise reward design based on taking into account learning feedback and addressing popularity bias, and proposes a Deep Q-Network recommendation system based on the psychological behavior to accomplish closed-loop psychological perception and personalized recommendation modeling. Experimental results show that the proposed model reaches an F1 score of 88.04% for psychological classification under the optimal temporal window. The precision of focused learning behavior recognition reaches 92.36%. Compared with baseline methods, there is a significant drop of 0.302 in the convergence loss, with improvements in training stability and model fitting ability. Visualization analysis reveals that this method has a dynamic response to changes in the psychological state of students during their learning process, which can optimize the adaptability of course recommendations as well as enhance the activities of online learning. The model provides technical support for learning state perception and adaptive teaching in intelligent online education.