PREDICTION OF AN EPIDEMIC SPREAD BASED ON ADAPTIVE GENETIC ALGORITHM. A CASE STUDY OF CORONA VIRUS

ABSTRACT This study develops an advanced Susceptible Vaccinated Exposed Infected Quarantined Recovered Fatal epidemic model to address limitations in traditional SEIR frameworks. The model explicitly incorporates vaccination dynamics (ν = 0.002), quarantine effects (κ = 0.05), and disease-induced fatalities (μ2 = 0.002), providing enhanced capability to analyze modern outbreaks where pharmaceutical and non-pharmaceutical interventions coexist. We implement a robust numerical solution using 4th-order Runge-Kutta integration with adaptive step-size control (hmin = 0.1 days), ensuring stability when handling the system’s stiffness during rapid epidemic transitions. The framework is paired with an Adaptive Generating Algorithm (AGA) that employs evolutionary computation to optimize high-dimensional parameters, demonstrating superior convergence compared to conventional estimation methods. Through extensive simulations calibrated to COVID-19 data, we validate the model’s predictive accuracy, achieving 94.2% agreement with observed epidemic curves (MAE=±127 cases/day). Sensitivity analyses quantify the relative importance of key parameters, revealing that transmission rate (β = 0.3) accounts for 41.3% of variation in R0, while quarantine rate (κ) contributes 32.7% to fatality reduction. The AGA optimization efficiently recovers parameters from noisy surveillance data, converging within 200 generations to solutions within ±5% of ground truth values. These results demonstrate the framework’s capability to handle real-world data imperfections while maintaining biological plausibility. The study yields critical policy insights, establishing quantitative thresholds for intervention effectiveness: maintaining vaccination rates (ν ≥ 0.005) delays epidemic peaks by 18–50 days, while activating enhanced quarantine (κ = 0.1) at 50 cases per 100,000 population reduces fatalities by 37%. We further demonstrate how waning immunity rates (ωV = 0.0005) and disease-induced mortality (μ2) shape long-term outbreak trajectories. The integrated RK4- AGA approach achieves computational efficiency 14× faster than Markov Chain Monte Carlo methods, enabling real-time scenario analysis. This work provides both theoretical advances in epidemic modeling and practical tools for outbreak response. The open-source implementation supports immediate deployment in diverse public health settings, while the modular architecture permits future extensions for agestratified populations or variant-specific parameters. By bridging mechanistic modeling with v