Final project in 42186: Model-based Machine Learning, spring 2021
Authors: Jonas Juhler-Nøttrup (S164217), Niels Asp Fuglsang (S164181), Morten Wehlast Jørgensen (S147056), Jens Thuren Lindahl (S124347)
In this project we investigated how Bayesian spatial count models can be used to model the number of youth pedestrian injuries in the census tracts of New York.
In order to run the full analysis and recreate the results, the notebook main-modelling.ipynb should be run. Note that the final experiments before uploading was conducted in Python 3.8 with a Macbook Pro M1, but running on different operating systems might result in slightly different simulations, hence, differing predictions.
The two datafiles are located in /data/. The code for loading the data (together with some preprocessing) is located in /src/utils.py.
The packages together with specific versions needed to run the code succesfully are listed in requirements.txt.
The notebook exploration.ipynb contains the initial explorative analysis that was conducted before handing in Milestone 1.