If I have a graph of customers and what products they've bought, Directed Edge makes it easy to figure out what products to recommend to each customer. Generating the same recommendations with Google's prediction API is not as obvious to me. What data do you train it with? The nth product the customer bought as the output, and the 1..n-1th products as the input?
Maybe I'm slow, but I don't see how google's prediction API is a good replacement for collaborative filtering type recommendation engines.
Comments
Directed Edge, a promising YC startup, makes recommendation engines surprisingly easy:
http://www.directededge.com/
It's quite a bit higher-level than what Google is offering here, with all the benefits and drawbacks that entails.
Thanks for the friendly plug :-)
You can find the full documentation on our developer site at http://developer.directededge.com/, and we offer a free developer account for non-commercial purposes: http://www.directededge.com/signup-developer.html
If anybody is giving the Google Prediction API a whirl for recommendations, we'd love to hear about your findings!
If I have a graph of customers and what products they've bought, Directed Edge makes it easy to figure out what products to recommend to each customer. Generating the same recommendations with Google's prediction API is not as obvious to me. What data do you train it with? The nth product the customer bought as the output, and the 1..n-1th products as the input?
Maybe I'm slow, but I don't see how google's prediction API is a good replacement for collaborative filtering type recommendation engines.