cool-pot / expo-mf-with-confidence

A Statistical Recommender Algorithm By Modeling Users' Exposure.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

ExpoMF with confidence

A modified version of expo-mf, which has a better performance than expo-mf when facing implicit feedbeck indicating "frequence or intensity".

This repository contains the experienmental study notebooks in my greduate thesis.

This repository is largely influenced by the paper "Modeling User Exposure in Recommendation" (WWW'16) and dawenl's repository"expomf".

Binarized Implicit Feedback

performance_steam

Non-binarized Implicit Feedback indicating "frequence or intensity"

performance_tps

Datasets

Contrast

I used the weighted matrix factorization (WMF) implementation in Implicit.

About

A Statistical Recommender Algorithm By Modeling Users' Exposure.


Languages

Language:Jupyter Notebook 48.1%Language:TeX 31.2%Language:Python 20.7%