Implementation of Collaborative Topic Modeling as in "Collaborative topic modeling for recommending scientific articles" by Wang and Blei (2011).
The model is used to create a recommender system for Steam games using the following datasets, also included in the repository:
- Steam 200k, containing information about per-game playtime for over 200k user-game interactions: https://www.kaggle.com/tamber/steam-video-games/version/1.
- Steam games complete dataset, containing various information for over 40k different Steam games, including title and description: https://www.kaggle.com/trolukovich/steam-games-complete-dataset
This repository contains:
- A Notebook.ipynb file that implements the model and showcases its performance against traditional Matrix Factorization.
- A ctm.py file for import convenience.
- The datasets used for the implementation.