<p>Analyzing high-dimensional data structures and interpreting them through visual strategies remain a significant challenge. Current methodologies often rely on time-consuming manual processes, such as interviews and intricate statistical analyses. In adolescent physical and mental health research, physical fitness tests and mental health surveys are primary assessment tools, employing questionnaires for mental health evaluations and fitness tests for physical activity indicators. Differences in analytical perspectives and data dimensionality contribute to variations in analytical outcomes, hindering comprehensive investigations and the development of targeted interventions. To address these issues, we introduce a multi-source data-oriented analysis method that uncovers associations across diverse data sources, reducing uncertainties in the analyses. Specifically, we leverage multi-source data and a self-organizing map (SOM) to reduce dimensionality for visualization, mapping sample points into a 2D topological space. K-means clustering is subsequently applied, using color and position variations to guide users' attention to the overall distribution of students' physical and mental health status, providing a comprehensive health outcome assessment. Parallel coordinates are designed to visually illustrate feature distributions, assisting users in selecting clusters of interest for analysis, effectively highlighting correlations across multi-source data. Furthermore, a personalized recommendation view based on Markov decision processes (MDP) is developed to support individual-level analysis. Integrating these data visualization and analysis functionalities, our approach offers effective tools for exploring and analyzing adolescent physical and mental health in depth. A thorough evaluation with real data and extensive user feedback demonstrates the high effectiveness and practicality of our proposed visualization method compared to existing approaches.</p>