This project is used as a capstone project for Eskwelabs. In this work, we developed models used to predict reported crime using different geospatial (light posts, train stations, etc.) and temporal (reported crime for the past week) features using geohashes and a weekly basis.
Run
conda env create -f crime-capstone.yml
to create a virtual environment.
You may also use the requirements.txt
to install the dependencies.
Notable dependencies include:
- numpy=1.16.4
- pandas=0.25.1
- matplotlib=3.1.0
- scikit-learn=0.21.2
- geopandas=0.4.1
- geohash=1.0
- data/: contains the files where the raw and preprocessed data is stored.
- notebooks/: contains the Jupyter notebooks used.
- pickle/: contains the pickle files for the models.
- source/: contains the class file for CrimePredictor and configuration files.
- utils/: contains utility methods like instantiating the model and loading the data.