4. As a design principle we specifically made sure that were significant inconsistencies in data captured and we didn’t try to correct the variations.
This was done to ensure that the system learnt over a period of time by having a feedback loop in the application to correct the data when it predicted a zone wrong.
To be practically viable, this also means that the system has to get thoroughly used and corrected in a supervised learning mode before being rolled out to users in real time.