Establishment of Ultrafine Dust Prediction Dashboard Prototype

• Research Objectives

  • Establishment of a dashboard prototype to effectively communicate ultrafine dust prediction results
  • Prediction of ultrafine dust based on data collected in real-time (every hour)
  • After comparing the prediction performance of each machine learning models, the prediction value and prediction performance of the best model are presented.

• Research Methods

  • Real-time air pollution data collection and preprocessing using Air Korea's Open API
  • Pre-train machine learning algorithms such as MLP, RNN, LSTM, CNN, etc., by optimizing calculation speed and using them for real-time predictions. 　

• Development of Ultrafine Dust Prediction Dashboard

  • Posting the webpage of the ultrafine dust dashboard app(http://keibigdata.github.io/Service.html)
  • Output of the best-performing prediction model. Providing hourly forecasts, data sources, and dashboard descriptions
  • We can check ultrafine dust forecast information and take countermeasures quickly.