Intelligent Evaluation and Prediction Model of Mental Health Status Based on Deep Learning

Abstract

In order to solve the problems of low efficiency and accuracy in traditional social psychological measurement techniques, the author proposes a deep learning based intelligent evaluation and prediction model for mental health status. The author combines multi parameter acquisition technology with deep learning algorithms and designs a psychological crisis testing algorithm based on a bipartite graph convolutional network model, using graph convolutional networks as the foundation. The algorithm is then embedded with a psychological testing instrument. The experimental results show that the accuracy of the BGCN undirected model is 88.52%, the accuracy is 86.21%, the recall is 66.56%, and the F1 value is 61.20%, all of which have performance advantages in model comparison. At the same time, with the increase of iteration times, the Loss change curve shows a stable downward trajectory, while the accuracy and F1 value curves show a large fluctuation amplitude in the early stage, a small fluctuation amplitude in the later stage, a fast decline speed in the early stage, and a stable trend in the later stage. This indicates that the model has the ability to conduct stable and accurate testing in the later stage of iteration. From the experimental results, it can be seen that this model can perform accurate psychological testing, which is conducive to the active development of social psychological testing.