3 ms·
Let's say you have the whole database of books from Amazon/GoodReads, including title/authors/genres/publishingdate/userrating/sales. You'd like to make a reco
by user5994461 6y ago
Let's say you have the whole database of books from Amazon/GoodReads, including title/authors/genres/publishingdate/userrating/sales.
You'd like to make a recommendation engine, the idea is that the user could input 1-3 books they liked and it would suggest more books that are similar.
What sort of algorithms should I look into to do that sort of things?
Note that I don't have user profiles with what book they read, only have the database of books, can't do the recommendation engine based on two users liked the same book so they will also like other books either of them liked.
- inoop 6y agoAWS has a service for that, it's called [personalize](https://aws.amazon.com/personalize/ https://aws.amazon.com/personalize/)
- meekrohprocess 6y agoIMO, you could highlight the differences between similar titles. It's easy for recommendations to get stuck on a local maxima if they only look at one metric at a time, like "similarity" weights. But if you have a lot of metadata about each title, you can break out of those "loops" by sprinkling in metrics like ratings/genres/release date/popularity/etc. This doesn't have to hurt from a performance perspective, either; you can filter on the same single metric, but request more recommendations than you need and pluck out a pseudo-random set in the application logic. That also lets you provide context for the recommendation. "It's like this, but [older/more obscure/with vampires]."
- hombre_fatal 6y agofwiw you’re asking how to build a reco engine without any of the necessary data needed to build one. Finding good data is the hard part. That’s the part that stops every HNer from building a reco engine as a hello world weekend project.