Research on Optimization of Intelligent Recognition Algorithm for Student Psychological Crisis and Construction of Graded Intervention Mechanism

Abstract

Early identification of psychological crisis risk remains challenging due to the complexity of behavioral patterns and the heterogeneous nature of educational data. This study proposes an intelligent risk-identification framework integrating temporal graph neural networks and multi-source data fusion. Behavioral information from academic activities, campus consumption, mobility trajectories, network usage, and textual records is collected to construct dynamic student profiles. A hybrid architecture combining graph convolutional networks, temporal sequence modeling, and self-supervised learning is developed to capture both social interaction patterns and temporal behavioral evolution. Experimental results demonstrate substantial improvements in identification accuracy, recall rate, and risk-classification consistency compared with conventional approaches. The framework provides an effective solution for intelligent risk assessment and behavioral analytics.