Abstract: In order to solve the problem of identifying the mental health status of college students, this paper based on students’ data of consumption, Internet access and mental status assessment result,firstly, Jenks Natural Breaks algorithm is used for feature classification, and then according to the results of feature classification, Apriori algorithm is used for feature association analysis to mine the behavioral features that have a certain correlation with students’ mental health status. Finally, the inertia weight is improved based on Particle swarm optimization algorithm, and the process of identifying, mutating and selecting inferior particles is added to avoid the algorithm falling into local optimal solution, and the firefly perturbation strategy is used to accelerate the convergence of particle swarm optimization to the global optimal solution. A PDNN (Particle Difference Neural Network) neural network model is constructed to predict the mental health state of students. The experimental results on the data set of students’ behavior characteristics show that the proposed model is superior to the traditional machine learning and related deep learning models, it can converge quickly and predict students’ mental health status more effectively and accurately. 

Key words:  , prediction of students’ mental health status, students’ behavior data, particle swarm optimization algorithm, identification and variation of inferior particles, neural network