Fine Dust Forecast Based on Recurrent Neural Networks

In this paper, we propose a fine dust forecast model based on deep neural networks. The proposed model uses five kinds of air quality-related information as input variables and presents fine dust levels on an hourly basis. For training, we built training datasets by crawling air quality open data provided by the Korea Meteorological Administration and Seoul City. According to the experimental results, the proposed method achieved an RMSE of 8.966 for the prediction of fine dust levels after one hour.