MONDEP: A unified SpatioTemporal MONitoring Framework for National DEPression Forecasting.

Area of Science:

• Mental Health Research
• Computational Epidemiology
• Data Science

Background:

• Depression is a widespread mental disorder with significant societal impact.
• Existing depression forecasting models often overlook administrative-level predictions.
• Untreated depression increases suicide risk, necessitating improved forecasting methods.

Purpose of the Study:

• To propose and evaluate a Spatio-Temporal Monitoring Framework for National Depression Forecasting (MONDEP).
• To develop real-time depression forecasting models using machine learning and deep learning.
• To predict depression prevalence at national and district levels using hierarchical aggregation and multivariate time series.

Main Methods:

• Data Pre-processing: Extraction and cleaning of raw national depression statistics.
• Exploratory Data Analysis (EDA): Visualization and analysis for data insights.
• Model Training and Testing: Application of machine learning and deep learning on multivariate time series data.

Main Results:

• The MONDEP framework effectively utilizes spatial-temporal data for depression forecasting.
• Deep learning models demonstrated superior performance, achieving a 13% improvement in Mean Absolute Error (MAE) over the SARIMAX baseline.
• The study confirmed a strong association between spatial-temporal components and depression profiles.

Conclusions:

• The MONDEP framework provides a robust method for national depression forecasting at various administrative levels.
• The findings highlight the potential of deep learning in enhancing the accuracy of mental health predictions.
• This framework can serve as a valuable tool for policymakers in timely mental health service management.