mdiby / fpmc

Implementation of the Factorizing Personalized Markov Chains for Next-Basket Recommendation

FPMC

Python/Cython implementation of the: "Factorizing Personalized Markov Chains for Next-Basket Recommendation" paper.

Notes

Only works for baskets of size=1. Shoulde be easy to change to other sizes. Our datasets only have these sized baskets.

Dependencies for library

• Cython
• Numpy
• Pandas

How to install

Clone the repo

::

$ git clone https://github.com/flaviovdf/fpmc.git

Make sure you have cython and numpy. If not run as root (or use your distros package manager)

::

$ pip install numpy

::

$ pip install Cython

Install

::

$ python setup.py install

Run the main script or the cross_val script:

$ python main.py data_file num_latent_factors model.h5

This will read the data_file, decompose with num_latent_factors and save the model under the filename model.h5

The model is a pandas HDFStore. Just read-it with:

::

import pandas as pd

pd.HDFStore('model.h5')

The keys of this store have the output matrices described in the paper.

Input Format

The input file should have this format:

dt user from to

That is, a tab separated file where the first column is the amount of time the user spent on from before going to to. The second column is the user id, the third is the from object, whereas the fourth is the destination to object. I used this input on other repositores, thus the main reason I kept it here. This code will ignore the first column, so you can just use any float.

References

.. [1] Rendle, S. and Freudenthaler, C. and Schmidt-Thieme, L. "Factorizing Personalized Markov Chains for Next-Basket Recommendation" - WWW 2010