Data‐Driven Deep Learning Framework for Ebola Transmission Dynamics Incorporating Deceased and Hospitalized Individuals With Real Data

ABSTRACT In this study, a nonlinear model for Ebola disease is developed incorporating transmission from both hospitalized and deceased individuals. A comprehensive mathematical analysis of the model is conducted. For a numerical solution, we design a robust neuro‐computing framework based on a deep neural network (DNN) approach. The standard fourth‐order Runge‐Kutta scheme is employed as a benchmark to generate high‐fidelity training data for DNN. The model is calibrated using weekly reported case data from Sierra Leone, one of the most severely affected countries during the 2015‐2016 Ebola outbreaks. Model parameters are estimated using least‐squares optimization via the least_squares function in the scipy.optimize module. Numerical results are validated through convergence analysis, error distribution, and regression‐based performance metrics under five cases using different sets of control parameters. The proposed methodology yields solutions that align precisely with benchmark data achieving a least absolute error of . The DNN architecture consists of three hidden layers with 10, 100, and 10 neurons each with the implementation of Tanh and ReLU activation functions. We believe that the integration of DNN with these activation functions in its hidden layers to study Ebola is a novel attempt at infectious disease modeling offering a powerful framework for epidemic prediction and control.