arushakelkar / yelp_dataset_recommendersystem Goto Github PK

PERSONALIZATION FINAL PROJECT

Course: E4571 Personalisation Theory, Fall 2019, Columbia University

Instructor: Prof. Brett Vintch

Team Members:

Arusha Kelkar ak4432 arushakelkar

Tanvi Pareek tgp2018 TanviPareek

Priyanka Lahoti pvl2111 PRIYANKALAHOTI10

Problem statement

Predict the last rating of an active user i.e. for user with 5 or more reviews, to hold out their final review (by date) and make a prediction on the rating of this final review.

Objective

The objective is to predict the last rating of each active user. Three models have been implemented and user-item bias baseline model have also been implemented to compare how well the models are predicting. The 3 models implemented are ALS, deep learning model using embedding layers and Factorization Machine(LightFM).

Dataset used : "Yelp dataset" The dataset can be downloaded from https://www.yelp.com/dataset/challenge

The report gives a thorough understanding of the models used and the results.

Requirements

• Python 3.7.4 with packages pandas, numpy, surprise, pyspark, keras, LightFM matplotlib installed

• 16 GB of RAM or Google Colab