Using Azure Machine Learning and high density WiFi to analyse customer behaviour
We’ve been involved in the delivery of complex network infrastructure for Microsoft for a number of years. One of the more challenging projects we regularly delivery is WiFi catering for several thousand concurrent users at Microsoft’s largest corporate conferences. Microsoft Ignite (formerly TechEd) and the Microsoft Australia Partner Conference have been delivered at the Gold Coast Convention and Exhibition Centre and cater for 3500 and 1500 delegates respectively.
Recent upgrades to the centre’s wireless infrastructure has increased the density of coverage such that the fine grained tracking of individual WiFi associations can be correlated against specific areas of the centre and the content being delivered in those areas. This in turn can be mapped back to the conference agenda to determine how effective specific sessions and speakers are. Finally, by crunching all of this data through Azure Machine Learning we can recommend content to users based on the inferential patterns of other users.
Privacy was a paramount consideration so all data was correlated by one-way hashed addresses so individual devices could not be correlated back to a specific person.
Other insights include, the ability to see the attrition of users in a given content stream vs other and also using machine learning to highlight broad patterns of behaviour such as a preference for individual one-on-one network as opposed to the consumption of session content.
The solution included ingestion of real-time radio association data from a Cisco WLC 5508 and approximately 250 radios via Azure Event Hubs. This data was streamed to SQL Server for historical reference and simultaneously delivered to an Azure Machine Learning catalogue for analysis. Finally PowerBI and a custom ASP.NET application provided Microsoft staff with an end user experience to visualise insights in real time.
For further information about this solution and how Codify can assist you with similar scenarios, please mail us at firstname.lastname@example.org.