back_to_portfolio
Illustration of mental health analytics using AI and machine learning

Research Paper · Submitted, Not Yet Accepted

Machine Learning-Based Mental Health Assessment and Depression Screening Using the U.S. DASS-42 Dataset

Co-Author · Submitted for peer review

PythonXGBoostRandom ForestANNUnder Review

Identifying depression early can help ensure timely intervention and improved mental health outcomes. This paper presents a machine learning approach to binary depression classification using a U.S.-specific subset of the publicly available DASS-42 dataset.

Three models were trained and compared under the same pre-processing conditions: Random Forest, XGBoost, and an Artificial Neural Network (ANN).

Infographic summarizing the dataset, methodology, model comparison, and impact of the depression classification study
Study overview — dataset (3,730 balanced samples), methodology, model comparison, and real-world impact.

XGBOOST

98.72%

Accuracy

ANN

97.20%

Accuracy

RANDOM FOREST

92.94%

Accuracy

Confusion matrices for XGBoost, ANN, and Random Forest
Fig. 2 — confusion matrices for all three models on the held-out test set.
Accuracy and loss learning curves for the ANN across training epochs
Fig. 3 — ANN accuracy/loss curves across training epochs, showing stable convergence.

XGBoost was the top performer and was further tested on unseen samples, where it proved suitable for fast prediction — making it a potential tool for digital depression screening for students and young adults.

This paper has been submitted for peer review and has not yet been accepted or published. This page will be updated with the venue and publication link once a decision is reached.