Enhancing Student Engagement in Online Learning Through Strategy Gradient Reinforcement Learning

At present, recommendation systems for online education often adopt a linear organizational structure, which leads to low recommendation efficiency and unstable response rates in the absence of human guidance from teachers. At the same time, personalized learning recommendation systems still cannot use analytical models to establish effective communication behaviors with learners, and students lack engagement. Given that deep reinforcement learning (DRL) introduces deep neural networks on the basis of reinforcement learning, as well as possessing the perception ability and decision-making ability of deep learning, it precisely fits the decision-making thinking mode of experienced teachers facing students with multiple levels of learning ability when setting questions. This article uses the deep deterministic policy gradient (DDPG) algorithm to construct a state environment, action set, reward function, and neural network structure to obtain the user’s cut set in the network graph based on user response records, which is the “learning area” topic for recommendation. It is a DRL algorithm that can handle continuous action spaces, combining the DPG algorithm and the AC framework. Through the experimental results, it can be concluded that the guided exercise recommendation model based on DDPG designed in this article can effectively establish communication behavior with learning users, and it achieve the requirements of guiding users to answer “learning area” questions through the learning of problem setting strategies, achieving good experimental results.