Skip to content

The hybrid model combining stacked denoising autoencoder with matrix factorization is applied, to predict the customer purchase behavior in the future month according to the purchase history and user information in the Santander dataset.

master
Switch branches/tags
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 

Collaborative-Deep-Learning-for-Recommender-Systems

The hybrid model combining stacked denoising autoencoder (SDAE) with matrix factorization (MF) is applied, to predict the customer purchase behavior in the future month according to the purchase history and user information in the Santander dataset. A blog post for some follow up discussions can be found here.

This work is contributed by Sampath Chanda, Suyin Wang and Xiaoou Zhang.

The user information is generated in "matrix_factorization.ipynb", and the rating marix is generated in "rating_matrix.py". The main code for the SDAE-MF hybrid model can be found in "mf_auto_mono.py" for one-hidden-layer SDAE and "mf_auto.py" for three-hidden-layer SDAE.

Update: in "mf_auto_mono_v2.py" and "mf_auto_v2.py", imporved the speed of matrix factorization by taking advantage of the sparsity of the rating matrix.

About

The hybrid model combining stacked denoising autoencoder with matrix factorization is applied, to predict the customer purchase behavior in the future month according to the purchase history and user information in the Santander dataset.

Topics

Resources

Releases

No releases published

Packages

No packages published