With the deepening of global population aging, the Elderly Care Service (ECS) industry has a growing demand for professional talents. The elderly care micro-major has gradually become an important way for universities to connect talent training with industry needs. However, the existing university-community collaborative curriculum system still has obvious deficiencies in the adaptability of curriculum content to industry needs, the matching degree of practical teaching with post abilities, and the curriculum optimization feedback mechanism. The system lacks a curriculum optimization mechanism that can realize data-driven dynamic adjustment. To address the above problems, this study proposes a university-community collaborative curriculum optimization framework for the elderly care micro-major based on Reinforcement Learning (RL). On this basis, an improved Deep Deterministic Policy Gradient (DDPG) model is constructed. By establishing a multi-objective reward function, the framework achieves collaborative optimization of curriculum content adjustment, practical arrangement matching, and teaching resource allocation. The study constructs an experimental dataset based on multi-source collaborative teaching data from 32 universities and 28 elderly care institutions. The dataset covers multiple dimensions, including industry post requirements, curriculum module information, student ability development data, and teaching resource constraints. The curriculum optimization strategy is trained and evaluated in an RL environment. Experimental results show that, compared with the traditional university-community collaborative curriculum model and other RL baseline models, the proposed optimization model achieves significant improvements in curriculum–industry adaptability, practical post matching degree, and talent training quality. Specifically, the curriculum-industry demand adaptability reaches 93.7%, an increase of 11.2% compared with the traditional model. The practical post matching accuracy rate reaches 89.5%, an increase of 9.8%. The student professional ability score increases by 23.6%, and the employer satisfaction rate reaches 91.3%. The ablation experiment results further demonstrate that combining the core optimization mechanism of RL with the university–community collaborative teaching model can improve the overall curriculum optimization effect by 15.4%. The research results indicate that the university-community collaborative curriculum optimization method based on RL can effectively improve the responsiveness of the curriculum system to industry needs. It also provides a data-driven and intelligent curriculum optimization path for talent training in the elderly care micro-major.
Meng et al. (Sun,) studied this question.