Welcome to RS_c, the central platform for the RecSys community. We provide curated lists of recommender-systems datasets, algorithms, books, conferences and many resources more. Maybe most importantly, we publish the latest recommender-system news. If you want your news to be reported on RS_c, read here.
An Exhaustive List of Methods to Evaluate Recommender Systems [Muffaddal Qutbuddin @TowardsDataScience]
June 18, 2020
Imagine we have built an item-based recommender system to recommend users movies based on their rating history. And now we want to asses how our model will perform. Is it any good at actually recommending users movies that they will like? Will it help users find new and exciting movies from plethora of movies available in our system? Will it help improve our business? To answers all these questions (and many others) we have to evaluate our model. Below I provide many different techniques to evaluate the recommender system.
First I will discuss the maths based methods for the evaluation. These help us lessen our algorithm options to use from bazillion algorithms out there. After that, I will touch on more business-related metrics to help choose the best technique for our business. In the end I will discuss a few real-life scenarios to help further our understanding of recommendation problems in real life and how it varies with the domain.