This is the source code repository for the Matrix Factorization (MF) implementation by Andrei Bondarenko and Geert Goemaere for the Project AI course (2021). We performed experiments on the Goodreads datset provided to us.
A small overview of the files in this repository:
- data: contains intermediate data files
- recommenders: contains the implementations of all recommendation algorithms (1st + 2nd presentation).
- sources.md: contains (some) of the sources we referenced
- utils.py: utility functions for notebook experiments
- preprocessing*.ipynb: preprocessing of the Goodreads dataset, different notebooks for different stages of the project (1st + 2nd presentation)
- baseline.ipynb: quantitative evaluation of the baseline popularity recommender (1st presentation)
- mf_grid_search_*.ipynb: grid search experiment for various MF recommender hyper-parameters (1st presentation)
- mf_qual_eval.ipynb: qualitative evaluation experiment for MF recommender (1st presentation)
- mf_quant_eval.ipynb: quantitative evaluation experiment for MF recommender (1st presentation)
- mfi_*.py: experiment scripts for implicit feedback matrix factorization (2nd presentation)
- mfi_*.ipynb: experiment notebooks for implicit feedback matrix factorization (2nd presentation)