Data Visualization in Support of Predictive Modeling of Childhood Depression with Machine Learning

This study investigates data visualization (DV) to support the construction of predictive models to diagnose childhood and adolescent depression. DV was used to guide decisions in pre-processing, such as cleaning, coding, and correlation analysis, and also in the interpretability of predictions using the SHAP technique. The analysis was applied to data from 378 individuals between the ages of 10 and 16. The results show that DV helped identify patterns, inconsistencies, and relevant variables, making the model more understandable and transparent. In addition, DV can assist clinical and technical decision-making by facilitating the interpretation of results, highlighting critical elements associated with symptoms of depression, and making the decision-making process more accessible, even in complex contexts. The study reinforces the importance of interpretable practices in using artificial intelligence in mental health.