Automating Analytics: How to learn metadata such that our buildings can learn from us

Automated analytics are the only way to cope with the big amount of data created by the internet of things in future. Good metadata such as the device type, sensor semantics, its location and functional context is essential to do this. But, in many cases metadata is missing or it is of poor quality itself. The paper explores what metadata is relevant, how it can be reconstructed and how it can be used to automatically deploy analytics at the example of Cognitive Buildings.