Multilevel Monte Carlo Simulation Model for Air Pollution Index Prediction of a Smart Network

Air pollution has become a significant environmental challenge in the 21st century due to widespread industrialization and urbanization worldwide. Effectively reducing it requires precise predictions of air quality. However, existing methods for predicting the Air Pollution Index (API) fail to effectively model short-term variables' dependencies and mostly neglect spatial correlations. Given these limitations, the statistical method emerges as the most suitable choice. Monte Carlo Simulation (MCS) is one of the best short-term time series prediction statistical approaches. This paper proposes a Multilevel Monte Carlo Simulation (MLMCS) model based on the MCS model for forecasting the API. The study covers the analysis of an air pollution dataset that includes the API and ambient air quality of ten locations in Beijing, China. The results show that the MLMCS improves the performance of API prediction compared to the MCS. The MLMCS model has the highest accuracy of 86.45% and the lowest computational time of 3.43 s compared to the MCS model's accuracy of 82.90% and computational time of 7.5 s.