Dynamic Psychological Risk Prediction for Students Using Fuzzy Inference and Deep Neural Networks

With the increasing severity of mental health problems among students, early warning and intervention for students’ psychological risks have become an important issue to be solved in the field of education. However, existing psychological risk assessment methods mostly rely on traditional scales and expert evaluations, lacking in-depth mining of students’ multidimensional emotional and behavioral data. To this end, this paper proposes a student psychological risk prediction model (FIS-DNN) that integrates fuzzy inference system (FIS) and deep neural network (DNN), aiming to improve the accuracy and realtime performance of psychological risk assessment. Based on the emotional and behavioral data of 800 freshmen to juniors in a certain university, this paper constructs the FIS-DNN model. Through five-fold cross validation, the performance of traditional DNN, fuzzy inference system and FIS-DNN models in student psychological risk prediction is evaluated respectively. The experimental results show that the FIS-DNN model is superior to the other two models in many indicators such as accuracy, precision, recall and F1 value. Especially in the case of dealing with situations with large emotional fluctuations of students, the prediction accuracy of FIS-DNN reaches $\mathbf{9 2. 3 \%}$.