Team:
Course:
- New York University
- DSGA-1004: Big Data
- Professor Brian McFee
Please see our final PDF report for a full overview of the recommender system and extentions developed.
Overview:
In the 21st century, many brick and mortar stores with limited inventory have been replaced by digital retailers with seemingly endless products and user-bases. Where physical retailers could only store popular items based on average user preferences, digital platforms have built recommender systems, which have become the primary way to interact with these large collections. The Goodreads dataset is one of these large datasets, containing over 800k users with 2.3M possible books to recommend. On average, users are associated with ~260 books, making selecting the next recommended book out of the remaining inventory a daunting task. This project seeks to build and evaluate a recommender system using explicit user feedback, in the form of a rating from zero to five, to recommend up to 500 books to users based on their preferences.
Exploration:
