<p>In vocational education, big data analysis of student behaviour has problems such as single data dimension, failure to effectively identify differences between student groups, and overly simple data analysis methods. Understanding student behavior in vocational education is essential for tailoring interventions that enhance learning outcomes, and leveraging big data enables more nuanced, data-driven insights that support precision education strategies. This paper applies the K-means clustering algorithm to deeply mine big data student behaviour patterns in vocational education, providing more precise and personalized data analysis tools and decision support for vocational education. The vocational education dataset, focusing on student behaviour data, is preprocessed through data cleaning, removal of invalid entries, and format unification. This is followed by behavioural analysis and key feature selection for clustering. The appropriate number of clusters is set, and the Euclidean distance is used to measure the similarity between each student and the centre point to complete the division of student behaviour patterns. The K-means clustering results are deeply analyzed to identify the behavioural characteristics of different student groups, extract valuable information, and analyze learning tendencies and participation, providing a basis for personalized education intervention. The experimental results show that the silhouette coefficient of K-means in 5 samples is between 0.78 and 0.92, and the variance ratio is between 0.82 and 0.91, which can achieve better clustering effects.</p>