Ranked 1st among 76 teams
The folder contains all the necessary code to create all the features and get the scores we had for submissions.
The repository is not cleaned yet and will be cleaned in the future.
Wolf Maxime, Palaric Aymeric, Levy Guillaume, Boulet Timothé
capital_countries.ipynb: creates the features like the country where the polygon is or distance to the nearest capial
convexity.py: contains a function that tests the convexity of a polygon (usef in preprocessing.py)
eval_model.ipynb: used for training of the model and predictions
extract_features_dates.ipynb: creation of the dates features (duration in days between today and the date, duration between 2 consecutive dates and duration to make an advancement between two status, etc.)
fourier_transform.ipynb: build fourier coefficients and fourier power as explained in the report
nearest_buildings.ipynb: contains 3 important functions that add the features of the kNN of the polygons, the mean of the features of the kNN polygons, and the area of the minimal polygon that contains the centroids of the kNN
preprocessing.py: performs basic preprocessing (see report)
utils.py: regroup all the contents of the other files to add all the features in the same dataframe, it is used to choose what features to add for the training of our model
other_models.ipynb: was used to create, train and compare different models
To generate the features: run
- preprocessing.py
- extract_features_dates.ipynb
- capital_countries.ipynb
- fourier_transform.ipynb
- nearest_buildings.ipynb
Then run:
- eval_model.ipynb (this will call load_data in utils.py to get all the features in the same dataframe)