Mental health problems of students in colleges and universities are becoming more and more obvious, and traditional assessment methods have limitations such as poor timeliness and narrow coverage. This study constructs a mental health assessment method for college students based on sentiment analysis technology and multilayer perceptual machine. The method mines the psychological state features implied in students’ web content data from two dimensions, text sentiment computing and image sentiment computing, respectively, through a multimodal fusion computing framework, and adopts a multilayer perceptual machine model for mental health level assessment. In the textual sentiment analysis part, the constructive study builds a three-layer neural network of word embedding layer-bidirectional long and short-term memory layer-dense connectivity layer to capture the textual contextual sentiment information; and in the image sentiment analysis part, a convolutional neural network based on VGG16 is used to accurately recognize the emotional tendency in the images through fine-tuning strategies. Aiming at the sample imbalance problem, the study introduces a cost-sensitive method to optimize the training process. The experimental results show that the evaluation method performs well on the sentiment classification MSD dataset, with an accuracy of 92.3% for text sentiment classification and 96.2% for image sentiment classification, both of which are better than the traditional CNN and BiLSTM models; and on the public dataset Yelp, the average accuracy reaches 67.08%, which is an improvement of more than 5% compared with other algorithms. The multilayer perceptual machine model is used for the student mental health assessment task with an accuracy of 92%, showing better generalization performance. The study shows that the multimodal fusion sentiment analysis technique combined with the multilayer perceptual machine model can effectively realize the automatic assessment of students’ mental health status in colleges and universities, providing a scientific basis for psychological intervention.
