The application of affective computing models in student psychoeducation and the optimal design of intervention pathways

In order to more intelligently and accurately analyze the mental health status of students through their emotions, and to promote the good development of students' mental health.In this paper, we propose a student mental health analysis method based on multimodal social emotion classification, using the multi-head self-attention mechanism in the BERT model to extract and train the textual features of the data, and then using the VGG16 model as a pre-training model to obtain the image features of the data, and then fusing the two features into a multimodal feature through the fully connected layer.The features are inputted into GRU layer and fully connected layer to get the subjective emotion of the student, and finally the dynamic matching image is emotionally categorized to get the side emotion of the student, and then the model of this paper is applied to the recommendation of psychological services for students.The student multimodal sentiment computation model improves its main accuracy by 2.86% compared to the better performing model in the comparison experiments.Psychological interventions for students based on the student multimodal emotion computational model led to significant improvements in students' total mental health, obsessive-compulsive symptoms, and other psychological factors, realizing the combination of psychological education work for college students with new technologies, advancing the progress of intelligent psychological work mechanisms, and enhancing the scientificity and relevance of psychological intervention mechanisms.