Recommendation System for Books
Click here to view the full notebook
Sample recommendation results
Harry Potter and the Sorcerer's Stone
Moby Dick
Angels & Demons
Charlie and the Chocolate Factory
Dataset
The dataset contains the 10,000 most rated books on Goodreads with their title, author, ISBN,
and nearly 6 million ratings of these books made by 50,000 users.
(Kaggle link)
Approach
Collaborate filtering makes book recommendations for a user based on how other users rate books.
Feature Engineering
The data is converted as a sparse matrix of ratings, in which each column is a user and each row is a book. Every entry in the matrix is a rating score.
Algorithm
kNN is used to find the n closest book vectors, with cosine similarity as the distance metric.