It is an important research direction of mental health discipline in the current era to evaluate and analyze college students’ mental health by using deep learning methods and form visual data characteristics and analyzable discipline conclusions. Based on this, this paper carries out the research method of convolutional neural network by using the research concept of deep learning. Firstly, the paper summarizes the fast intelligent analysis model based on the convolutional neural network system algorithm, classifies and summarizes the unique characteristics of college students’ mental health, and uses the convolutional neural network processing model to analyze, evaluate, and observe college students’ mental health combined with the big data theory. Secondly, through the expansion and utilization of multi-layer neuron self-coding neural network, the psychological health of college students is evaluated and analyzed in the psychological discipline, the discrete data structure is established by using the relevant psychological data, the psychological behavior of college students is analyzed, summarized, and classified, and the data model is filled to judge the mental health status of college students. Finally, through the design of confirmatory experiments, the results show that the college students’ mental health evaluation and analysis model based on deep learning is more efficient in individual data analysis. Compared with the mode of analyzing college students’ mental health through in-depth learning, the traditional psychological research method has a large workload and is not suitable for the universality and consistency of college students. This paper solves this problem and provides a reference for relevant research.

With the rapid development of modern society, there are many problems concerning the physical and mental health of students. This paper develops a feature analysis method of the mental health data of students in different colleges and regions and of different ages based on a convolutional neural network and TOPSIS evaluation model and studies the college students’ mental health analysis model based on convolutional neural network. First, through the data cluster summary and internal characteristics analysis of college students’ psychological questionnaire survey data in different regions and grades, we established a college students’ mental health grade system and evaluation index system. Then, the TOPSIS analysis method is used to analyze the characteristics of the data results, and the feasibility of the accuracy of the evaluation index standard is analyzed. Finally, the experimental results show that the college students’ mental health analysis model based on convolutional neural network can effectively classify and summarize various mental health data, quickly locate the mental health problems of different students and analyze the optimal solutions, and can effectively promote the process of analysis and research on the mental health problems in modern college students.