HYBRID ARIMAX-LSTM MODEL TO PREDICT AIR POLLUTION

Air pollution has a severe impact on public health and the environment. It is crucial to provide people with an air quality prediction tool to minimize health risks in the most affected areas. This study proposes a hybrid model that combines the auto-regressive integrated moving average with exogenous variables (ARIMAX) and long-short-term memory (LSTM) to predict air pollution levels. The ARIMAX model uses historical meteorological and air quality data, while the LSTM model captures complex nonlinear patterns in the residuals of ARIMAX predictions. With this approach, our hybrid model addresses the limitations of traditional and standalone models, achieving significant improvements in predictive accuracy. Experimental results show that our hybrid model outperforms individual ARIMAX and LSTM models, with a 31.9% improvement over ARIMAX and a 16.7% improvement over LSTM. This hybrid model offers a powerful tool for predicting air pollution levels that can be used in various domains, such as smart city applications.