Skip to content

yunduanduan/personalization

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Personalization Final Project

project-2-final-jlyc-fp created by GitHub Classroom

Team members: Xiao Ji ([email protected], xj2247), Xinyi Liu ([email protected], xl2904), Jiaying Chen ([email protected], jc5299), Duanyue Yun ([email protected], dy2400)

In this project, we used the Yelp dataset to build a recommendation system. We implemented 2 models: collective matrix factorization using the cmfrec package and factorization machine using the lightFM package. We compared the models against 2 baselines (bias and pure matrix factorization) and evaluated the models based on accuracy metrics and catalog coverage.

Please find the Final Report.ipynb for our final report.

Repository contents:

  • Codes folder
    • Bias baseline & user, item segmentation.ipynb: bias baseline model, user and item segmentation, build user attributes
    • CMF.ipynb: code for collective matrix factorization
    • Business.ipynb: exploratory analysis on business.json
    • MF baseline - ALS.ipynb: Matrix Factorization (using Spark ALS) baseline model
    • lightFM - Feature selection.ipynb: Factorization machine model feature selection
    • lightFM - cross validation.ipynb: Factorization machine model hyper-parameter cross validation
    • lightFM full dataset - overall results.ipynb: Factorization machine model full dataset precision and AUC
    • lightFM full dataset - precision by segment.ipynb: Factorization machine model full dataset precision by active/moderate/non-active users and popular/moderate/unpopular items
    • lightFM full dataset - auc by segment.ipynb: Factorization machine model full dataset AUC by active/moderate/non-active users and popular/moderate/unpopular items
  • Images folder
    • Images included in final report

About

Yelp Recommendation System

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published