Objective To explore the objective evaluation of depressive state of the youth in a certain unit with self-report scale combined with eye movement data, and to provide reference for the implementation of psychological intervention measures. Methods The Symptom Checklist 90 (SCL-90), Self-Rating Depression Scale (SDS) and Patient Health Questionnaire-9 (PHQ-9) were used to conduct psychological assessment on 200 young people from a certain unit from April 2023 to April 2024. According to the scores, depressive states were grouped, and the youth were divided into a depression group and a normal group. The eye movement data of subjects were collected, and the eye movement features were extracted. A depression classification model was established by using machine learning algorithm tools, and the classification accuracy of the model for a certain young people's depressive state screening was verified. Results A total of 200 questionnaires were distributed, and 169 were valid, with an effective rate of 84.5%.. By comparing the test results of SCL-90 with those of soldiers and local norms, the results showed that the scores of SCL-90 were lower than those of soldiers and local norms, and the difference was statistically significant (P＜0.01) . A population classification model was established by using eye movement data, and the classification accuracy of the youth depression was 79.29%. Conclusions The machine learning algorithm classification model based on the eye movement data of the subjects can initially realize the objective identification of high-risk groups of depressive states, and assist the traditional psychological assessment scale to identify depressive states.