Exploring the use of big data technology in mental health monitoring and intervention

Abstract In this paper, we take the mental health related data as the processing object, and construct a Chinese combined keyword extraction model that integrates mutual information, information entropy and D-S theory. The model and dictionary are combined with TF-IDF and SO-PMI algorithms to construct a special sentiment dictionary for the psychological domain, which improves the accuracy of mental health text analysis. The model and dictionary are used in the text analysis of mental health course discussions of college students, counting the high-frequency words of course discussion topics and analyzing the psychological problems implied by the high-frequency words. Comparison experiments are conducted to verify the accuracy and loss value advantage of the model in sentiment classification calculation. Implementing personalized intervention strategies for the analysis results, tracking the change of students’ mental health status before and after the intervention, and judging the application effect of the model. Analyzing the discussion texts of 400 college students in mental health courses, the model identifies high-frequency words such as “emotion”, and determines that the students’ mental health problems are caused by multiple pressures, and that they are accustomed to binge-drinking and overeating to digest their own problems. The model has a binary accuracy of more than 0.8 and a seven-dimensional accuracy of more than 0.5 in the correlation mining between emotion classification and mental problems, and the loss value is stable at about 0.16, with a high prediction accuracy. The intervention strategy based on the model and dictionary analysis results significantly reduced the students’ mental health SCL-90 scale scores (P<0.01), obtaining good monitoring and intervention effects.