Neural networks and behavioral frameworks to analyse the impact of media and individual awareness for controlling...

Enhancing individual awareness and immunization are generally recognized as fundamental strategies for controlling infectious diseases. Awareness is influenced by both mass media and interpersonal interactions, which affect individuals' protective behaviors and their readiness to participate in vaccination programs. This research establishes a hybrid framework combining evolutionary game theory (EGT) and deep neural networks (DNNs) to examine the influence of awareness on epidemic control. Classical epidemic models (ODE-based) often require comprehensive epidemiological datasets to estimate fixed parameters (e.g., infection rate, awareness impact, vaccine efficacy). In reality, however, such extensive data are not always accessible due to limited surveillance, reporting delays, or missing behavioral factors (such as awareness or media effects). Thus, here, we use the DNN framework not as a substitute for epidemiological data, but as a verification and approximation layer: (i) to confirm the consistency of our deterministic model, and (ii) to expand its applicability to scenarios when data is insufficient, or parameters fluctuate over time. Therefore, the proposed method utilizes EGT to encapsulate adaptive individual decision-making influenced by perceived payoffs, vaccination costs, and efficacy, while DNNs model epidemic trajectories and accommodate the nonlinear dynamics arising from awareness-behavior-disease feedback loops. More precisely, DNNs in epidemic models serve as a data-driven corrective layer that enhances deterministic ODE models, making them more flexible, more realistic (by reducing model-data mismatch), and better at capturing concealed social-behavioral feedback. Through numerical simulations, we present epidemic trajectories, two-dimensional heatmaps, and phase portraits to elucidate the co-evolution of awareness, vaccination, and disease transmission. The DNN component improves predicted accuracy by correlating simulated results with real epidemic curve patterns, thereby augmenting theoretical research with empirical confirmation. The findings indicate that effective vaccination initiatives, bolstered by targeted media campaigns and robust peer influence, significantly reduce the severity of the pandemic. The integration of EGT and DNNs underscores the significance of both behavioral adaptation and computational learning in epidemic forecasting and control. This combined effect provides a solid foundation for formulating adaptive policies, enabling policymakers to implement prompt, effective measures during epidemic or pandemic scenarios.