A Deep Learning Framework for Air Pollution Forecasting and Interpolation

Air pollution has been identified as the world's largest single environmental health risks by the World Health Organization. Real time air-quality information is necessary, to pretect humans against from the damage casused by air pollution. In this Thesis we will address this problem by creating a new framework capable of predicting and interpolating the PM2.5 concentration. We will use a Biderectional LSTM for the prediction part and an Artificial Neural Network with Self Training for the interpolation part. We will create 1km x 1km maps of the city of Chicago and we will compare our results with different baselines and existing frameworks.