AirNet: a machine learning dataset for air quality forecasting


Nov 03, 2017 (modified: Nov 03, 2017) ICLR 2018 Conference Blind Submission readers: everyone Show Bibtex
  • Abstract: In the past decade, many urban areas in China have suffered from serious air pollution problems, making air quality forecast techniques a hot spot. Conventional approaches rely on the numerical method to estimate the pollutant concentration and require lots of computing power. To solve this problem, we applied deep learning methods which have already achieved major breakthroughs in many other areas. Deep learning requires large-scale datasets to train an effective model. In this paper, we introduced a new dataset, entitled as ‘AirNet’, containing the 0.25 longitudinal and latitudinal degree grid map of mainland China, with more than two years of continued air quality measurement and meteorological data. We published this dataset as an open resource for machine learning researches and set up a baseline to a 5-day air pollution forecast. Through our experiments, it was demonstrated that this dataset could facilitate the development of new algorithms on forecasting the air quality.