A Hybrid Deep Learning Framework Based on CNN-GRU-TabNet for the Predictive Modeling of COVID-19 Mortality

The global outbreak of COVID-19 has presented substantial challenges in healthcare systems, demanding intelligent and responsive monitoring solutions. The integration of Internet of Things (IoT) technologies with Artificial Intelligence (AI) models has emerged as a promising approach to enable real-time surveillance and predictive healthcare. This study proposes an advanced hybrid deep learning model that combines Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), and TabNet for predicting COVID-19-related deaths using structured tabular data from India. The dataset comprises 4692 instances across 8 epidemiological features. The preprocessing involved mean imputation and normalization to handle missing values and scale the data. The CNN component extracts short-term temporal patterns, the GRU layer captures sequential dependencies, and TabNet applies attention-based feature refinement and selection. The model was evaluated using Mean Absolute Error (MAE), Median Absolute Error (MedAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and coefficient of determination (R²). The proposed CNN-GRU-TabNet model significantly outperformed traditional regression models, including Random Forest (RF), SVR, KNN, Gradient Boosting (GB), and Bayesian Ridge (BR), achieving an R² of 0.995 and the lowest error metrics. These results validate the effectiveness of the proposed hybrid framework for accurate and interpretable COVID-19 death prediction.