bgg_models is a project for training predictive models using data from boardgamegeek (bgg).
I train predictive models to estimate individual games at two levels: bgg community and bgg user.
Community models predict how the bgg community as a whole will evaluate games (average rating, number of user ratings, complexity) while user models estimate how individual users will evaluate games (likely to own).
My models are trained on game and collection data pulled via BoardGameGeek's API and stored via BigQuery.
This project requires:
- Authentication to GCP and BigQuery for data loaded from bgg.
- R (developed on version 4.3.0)
This project uses renv for package management and environment setup, and renv.lock can be used to restore the project state for running code.
_make_data creates a local copy of the data used in training the model
scripts that generate output (data and reports)
contains the source code used in jobs
Rmarkdown notebooks used in generating reports