Grid-based Feature Engineering

Aim

• To compute local spatial features to instill spatial awareness in the rocky terrain predictor.

• This implementation takes the point-based data as input, computes spatial features, and returns the dataset with spatial features.

Environment Setup

• Clone this repository.

git clone https://github.com/Gaurav0502/grid-based-spatial-feature-engineer.git

• Install all packages in the requirements.txt file.

pip install -r requirements.txt

• The scripts/spatialfeaturegenerator.py contains the code for generating the spatial features. To get the updated dataset (with the spatial features), you need to execute the compute_spatial_features() function inside the class SpatialFeatureGenerator.

• After execution, you must get four new features for every remote sensing index (i.e., any column except the depth and coordinate):

1. *_spatial_mean
2. *_spatial_std
3. *_spatial_weighted_mean (excludes the point itself*)
4. *_spatial_weighted_std (excludes the point itself*)

* prevent zero division error.

Notes:

• The function only computes features for a grid of dimensions (3 x 3). To use any other dimensions, you must use the compute_spatial_features_with_stride() with the grid_size as input and the stride for the sliding window.

• Any values outside the bounds of the feature space are considered as NaN and are excluded from the computation.

• When using the output dataframe for modelling, ensure you make use of the GroupKFold from sklearn to mitigate chances of data leakage. In other words, a grid (in its entirety) must either be in the train set or the test set.

• Preferably, keep the grids for both depths in the same set to instill spatial awareness and depth perception into the model.