FancyYelpers: Yelp Restaurant Recommender System

NYU Data Science 1001 Project, Fall 2020

Data:

Yelp Open Dataset (https://www.yelp.com/dataset)

Overview:

The advent of crowd-sourcing review platforms makes business information readily accessible for customers. Yet, as the number of choices becomes overwhelming, there is need to filter, prioritize and personalize relevant information in order to alleviate the information overload. In this paper, we build a Yelp restaurant recommender system to provide users with personalized restaurant recommendations. We build popular collaborative filtering (CF) and content-based (CB) methods which consider user-business interactions, restaurant attributes and text mining of user reviews. We also construct the state-of-the-art LightFM hybrid model which unites the advantages of content-based and collaborative filtering recommenders and thus significantly improves the predictive power. Our framework to build the outperforming recommenders serves as a practical guide to help the platform provide commercial restaurant recommendation service.

Models:

• Collaborative Filtering via Matrix Factorization
• Content-Based Filtering
• Model Ensemble
• LightFM Hybrid Model

Methods Used:

• Data Visualization
• Model-Based Collaborative Filtering
• Matrix Factorization
• Natural Language Processing
• Content-Based Filtering
• Cosine Similarity
• Latent Semantic Indexing
• Model Ensemble
• Hybrid Model

Final Report

The final report can be found here

Team Members:

• Si (Steve) Gao (si.gao@nyu.edu)
• Xinli Gu (xg588@nyu.edu)
• Xirui Fu (xf620@nyu.edu)
• Wenxin Zhang (wz2164@nyu.edu)

Questions?

Please feel free to contact me should you have any questions.