A Simulation-Based Approach for COVID-19 Vaccination Strategies: Integrating Clinical Features for Prediction and Outcome Analysis

Abstract

Global health has been significantly changed by the COVID-19 pandemic, which demands the development of efficient vaccination strategies for counteracting the virus spread. A simulation-based approach that combines clinical features to enable vaccination outcome prediction and strategy optimization is presented in this study. Key features such as age, pre-existing health conditions, immune response likelihood, exposure risk and vaccine type were modelled using synthetic data generated from around 1,000 patient profiles. Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) were employed to study the vaccination strategies. The CNN model achieved 96% training accuracy and 94% validation accuracy, while the RNN model achieved 95% training accuracy and 93% validation accuracy. CNN and RNN reached ROC AUC values of 0.93 and 0.92 respectively, indicating strong predictive capability. The study was further extended to explore the impact of regularization techniques such as dropout (30% and 50%) and batch normalization; accuracy increased by up to 2.5% and loss decreased by 1.5%. Training ran for 50 epochs with early stopping to prevent overfitting. The results suggest that demographic and clinical factors can inform personalized vaccination strategies that improve public health decision making. Limitations include the use of synthetic data and the need for real-world validation. Future work will extend the datasets and add clinical variables towards robust and generalizable models.