Fractional-LSTM hybrid modeling for Ebola transmission prediction

The transmission dynamics of Ebola virus disease are investigated using a fractional-order compartmental model combined with temporal prediction techniques. The population is divided into susceptible, vaccinated, exposed, infected, critical, hospitalized, funeral-associated, recovered, healthcare-worker, and environmental compartments. A variable-order fractional formulation is employed to incorporate memory effects in disease progression. The system is solved numerically using the Predictor-Evaluate-Correct-Evaluate (PECE) method for different fractional orders, revealing distinct epidemic behaviors. To enhance prediction accuracy, the numerical time-series outputs are used as input to a Long Short-Term Memory (LSTM) network, which learns residual variations and produces corrected epidemic trajectories. The proposed hybrid framework preserves the epidemiological structure while improving smoothness and predictive reliability. Develops a variable-order fractional compartmental model capturing memory effects in Ebola transmission across multiple epidemiological states. Integrates numerical PECE method with LSTM-based residual learning to refine epidemic trajectory predictions and enhance accuracy. Proposes a hybrid physics-informed and data-driven framework that preserves model interpretability while improving smoothness and predictive reliability.