Exploring the reform of college mental health education based on multi-source data

Abstract

Abstract With the primary study emphasis on feature extraction, this article first suggests an algorithm for identifying students’ mental health issues based on multi-source data and employs one-dimensional convolutional neural networks to extract the online trajectory patterns of pupils from online behavior segments. Anomaly ratings were computed based on student consumption statistics from the cafeteria in order to depict the dietary variations among students. Next, a DeepPsy-based algorithm for identifying pupils with mental health issues was recommended after the algorithm suggested in the preceding phase was improved. A two-dimensional convolutional neural network was used to build the Internet trajectory matrix and retrieve the daily Internet trajectory pattern. A long and short-term memory network was used to record the time-dependent connection between each day, and a deep learning network was created by fusing the fundamental features and the Internet trajectory pattern. Through experimental research, the DeepPsy model was able to detect 75.5% of pupils with mental health issues and obtained a Recall of 0.755, outperforming the multi-source-based algorithm by 19.5%. Obsessive-compulsive symptoms took the top spot among the many variables influencing mental health, coming in at 22.8%, followed by interpersonal sensitivity at 17.25%. Depression came in third with a proportion of 10.76%. As a result, instructors can use the algorithm suggested in this article to spot pupils who may be suffering from undiagnosed mental health issues.