On July 1st 2015 I visited the NoSQL Usergroup in Cologne to give a session on how to build recommendation systems in Azure.
I went through various demos highlighting the three types of recommender engines (collaborative filtering, content-based filtering and hybrid recommender):
- The simplest of Machine Learning problems: using binary classification for targeted marketing, i.e. which customers is it worth to send a marketing mail?
- Create association rules using R in RStudio and then integrate in AzureML
- MatchBox Recommender already included in AzureML
- Frequently Bought Together plugin from the Azure Marketplace
The first three experiments then can easily be deployed as a web service within seconds/minutes.
The accompanying slide deck is the following one:
Last but not least: Special thanks to my colleague Alexei Khalyako who provided the basis for this session. And thanks to the NoSQL Usergroup in Cologne for hosting me! :)