Recommendation engines are all over the web. We find them when we purchase things at Amazon, when we watch movies from Netflix and when we add new Friends to Facebook. The idea behind these recommendation engines is that based on similarities of items that are found between you and other users of the same application, the engine can infer your probable likeness of other items that you have not tried, books you haven’t purchased, movies you haven’t watched, friends you haven’t friended.
At Intela we are always pushing the limits of what the technology can do for our systems and applications and we have been working on ways to improve how we deliver a better service to our customers.
So a few months ago we started a project to look for a good recommendation engine that would allow us to create better and more targeted campaigns by using recommendations based on past behaviour from our users. We looked at several technolgies and concentrated our efforts on 2: The Google Prediction API and Apache Mahout.
We started working on a proof-of-concept application using the Google Prediction API which was in Beta at the time and built a very simple application that would give us a good idea of what the capabilities of the API were. We worked on several variations of it but in the end we found it was too restrictive for our needs and decided to move on to the next one.
Apache Mahout is an open source project that contains an impressive library of machine learning algorithms including recommendation engines, classification, clustering data mining and many other algorithms. We concentrated on creating once again a proof-of-concept for a recommendation engine. This time it took a little longer to familiarize ourselves with the libraries and API and how to provide the data for the recommender, but we soon had a very good application that was delivering the results we were expecting.
After a few months working on this we have finally deployed a couple of implementations of our own recommendation engine based on Mahout and we are starting to see very good results.
Intela has a reputation for very inteligent marketing… ladies and gentlemen… Intela has just gotten smarter.
Posted by: Emilio Suarez, Senior Software Architect





Ryan Wilson, SVP Intela republished from 

