Research on data mining and pattern recognition of college Chinese learning behavior in higher vocational colleges for teaching optimization

Abstract

This study targets insufficient data resources and backward teaching modes in higher vocational Chinese education. It builds a three-dimensional data collection matrix covering classrooms, virtual communities and enterprise practice, gathering 12 data types from 2,000 students of five majors. OpenPose captures students’ head posture angles to generate an attention index for evaluating learning engagement. Combining text semantic analysis and visual behavior data, the research constructs heterogeneous information network feature maps and adopts graph convolution network (GCN) to optimize node embedding, matching learning behaviors with competence requirements. It also applies multi-task learning and dual Q-network to build a personalized teaching intervention model based on Markov decision processes (MDP. Experiments prove the approach works well: the experimental group sees a 27.6% rise in ability mastery, teachers’ decision time drops to 3.2 minutes, and students’ task completion rate reaches 89.7%. This study improves traditional educational data mining and offers a replicable model for vocational teaching optimization via multi-modal behavioral data.