Yelp Rating Prediction with and without Graph Embeddings from Neo4j

Exploring machine learning and graph embedding concepts to build model to predict user rating for various restaurants from yelp data

Folders

• Dataset - download yelp dataset from (https://www.yelp.com/dataset) and save users.json, business.json and reviews.json files here
• Neo4j dataset - files from "Build relevant yelp relational database for Neo4j" are saved here
• results - screenshots from Machine Learning model results and Neo4j

Notebooks

• MLPNN_rating_prediction - MLPNN model trained on top 10 restaurant category yelp data for rating prediction - dataframe heads are printed to show data at multiple stages.
• RF_rating_prediction - Random forest model trained on top 10 restaurant category yelp data for rating prediction
• RF_with_embeddings_rating_prediction - Random forest model trained on top 10 restaurant category yelp data along with node embeddings obtained for restaurant categories for rating prediction
• Build relevant yelp relational database for Neo4j - Generates relational database files (json format) from top 10 restaurant yelp data - used to build graph databse in Neo4j - files are saved in Neo4j dataset folder -> they have to be later moved to Neo4j relevant import library foler when a database is initialized in Neo4j
• Neo4j_cypher_queries - cypher queries used to load json files and create yelp graph database in Neo4j with node and edge properties + queries to extract graph embeddings.

Misc

• MLPNN_rating_prediction_with_dataframe_headers.pdf
• MLPNN_rating_prediction_with_dataframe_headers.html