Building a Book recommender system using Popularity-based filtering, collaborative filtering and content-based filtering.
The datasets are from bookcrossing.com, a free online book club which was founded to encourage the practice of "leaving a book in a public place to be picked up and read by others, who then do likewise", aiming to "make the whole world a library." The dataset was collected during year 2004.
- It is a subclass of information filtering, that seeks to predict the "rating" a user will give to an item, and basis these predicted ratings, recommends Top-N rated items to that /user
- It helps users by suggesting them most relevant items from a large corpora.
- In other words, it is a useful alternative to search algorithms, as they help users discover items they might not have found otherwise
- Increases Revenue by directing the items and sales offers to specific users, and thus increasing the likelihood of selling items.
- Increases Customer retention: It helps business in understanding the needs of customers & eventually, in devising the right strategies to foster and maintain relationship with them.
- Makes User’s life easier by handling a large amount of information and by providing them with personalized, exclusive content and service recommendations, which further leads to customer satisfaction.
- Majority of readers are from USA, followed by Canada.
- The average of ratings given by Kids was the highest, followed by seniors; On average, rating given by readers from Spain & France were the highest.
- For new users, built global popularity based recommender systems.
- For registered users with no rating history, built demographics based popularity recommender systems.
- For users with ratings history, built collaborative, content based and hybrid recommender systems.
- For this dataset, Content-based recommender has the highest catalogue coverage for all N (as, ~43% for N=10), while Hybrid recommender system has the highest MAP@N for all N (as, MAP@10 = 2.43%).