Abstract
In recent decades, the air quality issue has caught everyone’s attention and become a significant problem for everybody. It has influenced human living and brought a variety of risks to the health of people. However, it is always difficult to balance air quality and economic development. In this study, we consider the recent debate on energy policy making and investigate the forecast of air pollution in Central Taiwan. Based on long-term time-series past data of PM2.5 and relevant chemical and meteorological factors, we use several different types of popular machine learning approaches for predicting the concentration levels of PM2.5. In particular, the autoregressive hidden Markov model (AR-HMM), which admits the existence of dependency between time-series observations, has a relatively better prediction performance. The empirical studies in Taichung area, Taiwan demonstrate the effectiveness of the model, which can be used to assist the government for setting appropriate energy policies.