Despite the significance of vocational education in an AI setting, the issue of erroneous research persists. Not only does the classic particle swarm method not work in an AI setting, but it also fails to satisfactorily tackle the issue of vocational education. Based on an adaptive learning recommendation system, this study analyses the current state of vocational education and recommends new directions for it within the AI framework. In order to lessen the impact of outside forces, we first utilize the theory of user behavior and preference to identify the elements that have an impact, and then we categorize the indicators according to the needs of vocational education. Next, an adaptive learning recommendation system for vocational education is developed using user behavior and preference theory. The outcomes of this algorithm are then thoroughly examined. Based on certain assessment criteria, the MATLAB simulation results demonstrate that the adaptive learning recommendation algorithm outperforms the classic particle swarm approach when it comes to vocational education impacting factor time and accuracy.