README.md

Steps to follow for recommendation on Yelp dataset

a) Pre-run step:

• Clone the repository
• Install packages and dependencies: pip install -r requirements.txt
• Download Yelp dataset from here: https://www.yelp.com/dataset/challenge
• Place the extracted folder into the data/ of the repository
• Run Data_Pre_Processing.ipynb from src folder. This will create the relevant datasets necessary to run the models we have defined

b) For Bias Baseline and ALS model:

• Run src/ALS_Baseline.ipynb

c) For Factorization Machine model:

• Run src/CMF_FM.ipynb

d) For Wide and Deep model:

• Run src/Wide and Deep.ipynb

e) For Content-based recommendation:

• Download glove, infersent model mentioned here: https://github.com/facebookresearch/InferSent
• Move the infersent2.pkl to src/Content-Recommendation
• Run python src/json_to_csv.py to convert json to csv consisting of Las Vegas restaurant dataset for 2018.
• Run python src/Content-Recommendation/get_review_embedding.py to generate weighted review2vec and export it to a file.
• Run python src/Content-Recommendation/content_recommendation.py to create annoy index from review embeddings and provide top 10 recommendations for the input string.

Models Used:

• Wide and Deep Recommendation System

• Infersent

• Universal Sentence Encoder

• KGAT: Knowledge Graph Attention Network for Recommendation

• Graph Embedding Based Hybrid Social Recommendation System

• Alternating Least Square

• Factor Machines

Team

L-R: Benjamin, Siddhant, Swarna and Amogh