This project seeks to create a book recommendation model from the Book-Crossing dataset (available here).

Building a recommendation system: full workflow - github repo here.

Implementing recommender sys from scratch in Python.

Recommender system library: suprise (see here)

Intro to recommender systems using collaborative filtering.

The Book-Crossing dataset is comprised of 3 tables:

1. BX-Users
   This contains the Book-Crossing users. There is a 'User-ID', and some demographic data ('Location', 'Age') if available.

2. BX-Books
   Books are identified by their ISBN number, and invalid ISBNs have already been removed from the dataset. Where available, book information is given here ('Book-Title', 'Book-Author', 'Year-Of-Publication', 'Publisher'). If there are several authors for one book, only the first is given. URLs linking to cover images are also given in small, medium, and large formats ('Image-URL-S', 'Image-URL-M', 'Image-URL-M').

3. BX-Book-Ratings
   Contains all of the book rating info. Ratings ('Book-Rating') are either EXPLICIT (on a scale of 1-10 - higher being better), or IMPLICIT, which is expressed by a 0. i.e. a ZERO is NOT a bad rating, but just an implicit one.

The datasets were downloaded as a csv dump, where the first line contains column names and fields are separated by a semi-colon.