Seoul Urban Data Science
In this project, we researched factors that impact on air quality levels. First, we collect millions of records of historical data including weather data and air pollutant data from 2010 to 2018 of many cities in China and South Korea. We evaluated the accuracy of different deep learning methods on air quality forecasting for these cities. We published three papers in this project, which present many valuable findings.
Our earliest publication in this project presented a simple deep learning approach to air quality forecasting based on LSTM network, which can be found here. This paper has been cited by many research papers in air quality forecasting.
As filling missing data is a crucial process to build high-accuracy forecasting models, our second publication aimed at addressing data interpolation and air quality forecasting, which can be found here.
Later on, we proposed a multi-modal approach that incorporate multiple deep learning models to study underlying patterns in air quality variations. In this paper, we tried to assess the impacts of different factors such as weather, air pollutants, and trans-boundary air quality levels on the future air quality levels. The link to this paper is here.