With the popularisation of education, the number of college students is increasing day by day, and there are also more students with psychological health problems. Whether students’ psychological abnormalities can be detected in time is one of the main problems faced by colleges and universities at present. Adopting digital technology to mine, collect and analyse the data generated by psychological health education in colleges can effectively solve the dynamic development of students’ psychological health problems. Therefore, in this paper, the psychological health problems of college students are identified and classified by establishing an improved logistic regression model. The behaviour characteristics are quantified and the differences are combined according to students’ relationships with their classmates, life rules and economic conditions. The test results show that the regression effect of the model is excellent, which can identify college students’ psychological health problems and improve the intervention and treatment of educators on students’ psychological problems.