Improved Recommendation Algorithm by Integrated Clustering in the Construction of Intelligent Learning Model in Higher Vocational Colleges

In the context of the information technology and big data era, higher vocational colleges are faced with the dual challenges and opportunities of data utilization and innovation in teaching models. Traditional teaching methods struggle to fully exploit the value of student behavioral data, and recommendation algorithms have limitations when handling complex student behavior data, making it difficult to meet personalized learning needs. This study aims to optimize recommendation algorithms using ensemble clustering techniques, constructing an intelligent learning model tailored to higher vocational colleges to enhance student engagement and participation. The innovation lies in two key aspects: first, a learning recommendation model based on Dynamic Clustering (DC) and Deep Reinforcement Learning (DRL) is proposed. Initially, traditional K-means clustering is employed to perform a preliminary clustering analysis of student behavior data, which provides an initial segmentation of the student group. The DC algorithm is then introduced to model and analyze student behavior data dynamically over time, accurately capturing trends and patterns in students' learning behaviors. Second, the DRL algorithm maps student behavior data to the learning recommendation decision space. Through interactions between students and the environment, the model learns to optimize recommendation strategies. Furthermore, ensemble methods are used to combine multiple model predictions, effectively reducing bias and variance, and enhancing both model generalization and recommendation accuracy. This study trains and evaluates the proposed learning recommendation model using the Student Classroom Behavior (SCB)-Dataset3 dataset. The model’s effectiveness is quantitatively assessed using objective metrics such as course view counts, click-through rates, viewing duration, attendance rates, learning activity indices, and resource utilization rates. The results show that the designed intelligent learning model demonstrates significant advantages on public datasets. Compared to traditional methods, the accuracy and personalization of course recommendations have notably improved. Specifically, course view counts increased by 10%, click-through rates improved by 8%, viewing duration rose by 12%, attendance rates increased by 5%, learning activity indices grew by 15%, and resource utilization rates rose by 30%. The integration of clustering techniques and recommendation algorithms in the intelligent learning model proves to be both feasible and superior in higher vocational colleges, significantly enhancing students' learning experience and outcomes. The findings of this study provide reliable intelligent learning models for higher vocational colleges, supporting personalized learning recommendations, which have important practical significance and promotion value for improving teaching quality and student satisfaction.