Cloud based real-time collaborative filtering for item–item recommendations

• Proposal of a new general architecture for a real-time collaborative filtering system that produces item–item recommendations.
• The proposed architecture makes use of all available input data, as opposed to data sampling, to produce relevant recommendations.
• It also makes use of on-demand computing resources (cloud computing) to scale up and down, and adapt to variations in the number of items as well as users.
• Shows a real industry application of the architecture.

We describe a large scale implementation of a video recommendation system in use by the largest media group in Latin America. Taking advantage of existing recommendation system techniques, the proposed architecture goes beyond the state of the art by making use of a commercial cloud computing platform to provide scalability, reduce costs and, more importantly, response times. We discuss the implementation in detail, in particular the design of cloud based features. We also provide a comprehensive generalization of the architecture that allows its application in other settings.