Music recommendation supported by AI algorithms has been widely used in college students’ mental health interventions. In this study, we constructed a three-dimensional entity network of “college students-music resources-socialized labels” by mining and analyzing the features related to the social labels of college students on music social networking platforms to correlate college students’ interests in music resources. Multidimensional Correspondence Analysis (MCA) is used to reduce the noise of heterogeneous tag data, and the Boolean matrix of music and tags is established to improve the correlation between tags and resources. Combining the music interest degree scores of college students with the music interest similarity calculation results of neighboring college students, sorting the music to be recommended, recommending music for college students that matches their psychological state, and designing controlled experiments of music intervention for mental health. Through the systematic music recommendation to intervene in the mental health of college students, the four assessment dimension scores of students in the experimental group after the experiment had an improvement of 27.560±3.425 to 37.040±2.613 points.
