chahbaz-aman / blueskyabove_hackathon

Submission for the Blue Sky Above Challenge conducted by IEEE.

BlueSkyAbove_Hackathon

Submission for the Blue Sky Above Challenge conducted by IEEE.
https://www.hackerearth.com/challenges/hackathon/ieee-machine-learning-hackathon/

Project team:

Ditipriya Gorai
Chahbaz Aman

Running Procedure:

1. Clone the repository.
   This is necessary to download/save large satellite data files in the data/L2 folder.
2. Run the BlueSkyAbove_Solution.ipynb notebook.

Solution Overview:

The model implements a Random Forest Regressor as the NO2 estimator.
7 days prior satellite data is stored in data/L2. For a specified time and place for NO2 estimation, the most recent reliable satellite data is searched
and extracted from the data files. This 2D spatial data is compressed into a 1D array of float values fed to the regressor. Further, map data from OpenStreetMap for the place is downloaded and features are extracted from the image in terms of land-use scores. The regressor uses the satellite data, land-use scores and time data to estimate NO2 concentration in the air.

Figure: Solution Overview