Perception issues in data mining

Perception is crucial to data mining. When dealing with real-world problems, e.g., expert systems analyzing stock market and financial data, etc., there is a difference between the real world and what the user (or expert system) perceives to be the real world. Often data mining retrieves perceived data and the problem is to reconcile this perceived data with the real world. For example, data mining might retrieve a perceived set of rules learned from long experience by a plant operator in operating a plant (and which can vary significantly from the mathematical model of the plant system). How do we reconcile the two different models? Which is the real one? Information granulation can help with this problem.