Enhancing Epidemic Prediction Using Simulated Annealing for Parameter Optimization in Infection Network Inference

Understanding and predicting outbreaks of epidemics has become a major focus since COVID-19. Researchers have explored various methods, from basic curve fitting to complex machine learning techniques, to predict how the virus spreads. One promising method is the Network Inference-based Prediction Algorithm (NIPA), which uses the SIR-model and the least absolute shrinkage and selection operator to estimate how the infections spread over different regions. However, fine-tuning the regularization parameter of NIPA can be complicated because of the time-consuming process and sub-optimal result of k-fold Cross-Validation (CV). To overcome this, we suggest using Simulated Annealing (SA) to optimize NIPA's regularization parameter and find an optimal value for the curing probability. Our study aims to combine SA with NIPA to make the process of choosing the optimal value for the parameters more effective. The results of the research show that the accuracy is improved and therefore indicate that SA is an acceptable alternative to CV, regardless of the computation time being higher.