AAI_2025_Capstone_Chronicles_Combined

MENTAL HEALTH RISK DETECTION USING ML​

​

​

​

​

10

Several studies demonstrate the practical value of the selected models in the contexts of mental health and clinical prediction. For instance, Logistic Regression was successfully applied in COVID-19 mental health screening, where it modeled psychological instability based on stress and isolation features (Alkhamees et al., 2021). Similar approaches were used in cardiac mental health assessments to predict depression and anxiety using lifestyle indicators. Tabular Neural Networks, as shown in Somepalli et al. (2021), outperformed traditional models in disease onset prediction, showcasing their ability to model complex feature interactions. Additionally, personalized mental health monitoring systems have used neural networks to track risk changes over time based on behavioral inputs. Finally, Soft Voting Ensembles have been used in psychiatric diagnostic systems to combine Logistic Regression and deep learning outputs, improving both sensitivity and classification accuracy (Jain et al., 2025). These examples, summarized in Table 3, validate our model choices and demonstrate alignment with state-of-the-art techniques in applied machine learning.

Table 2 Similar Projects Using These Methods

Method

Project / Study

Description & Relevance

Logistic Regression

COVID-19 on Psychological Stability (Alkhamees et al., 2021) Cardiac Mental Health Risk Prediction

Predicted psychological risk during quarantine using stress and isolation variables

Identified depression and anxiety risk in cardiac patients using lifestyle and sleep features.

Tabular

Deep Learning on

Demonstrated superior performance over tree-based

210