sjtuzhou / netflixprizematrixfactorization Goto Github PK

Netflix Prize

Implementation of Matrix Factorization from the article

https://datajobs.com/data-science-repo/Recommender-Systems-[Netflix].pdf

Use algorithm 'stochastic gradient descent' and 'alternating least square' to train parameter matrices,

Use multi-processing in python by default

const.py: Specify the path and filename constants. You can change to your local file path.

prepare_data.py: Run it first to get the dataset in format of NumPy and other auxiliary files. It will divide train set and test set randomly.

ALS_extra_data.py: Run it AFTER prepare_data.py to get the extra needed data in format of NumPy for ALS method

matrix_stoc_grad_desc.py: Run it to train and evaluate the matrices by stochastic gradient descent

ALS_matrix.py: Run it to train and evaluate the matrices by alternating least squares

All files and directories related to keras is a test version.