I'm trying to label customer data with a word describing how typical they are. There is basically 3 possible values: typical, temporarily untypical, untypical. But I'm not sure if "untypical" is the correct word. Should it rather be "atypical" or "nontypical"? Does it matter if we're talking about technical things? Is there a difference for british and american english?
Considering the recent activity and interest, I'll give an explanation of what we're doing. We're trying to predict customer behaviour (product consumption per day). To do this, we have a long history of measured product consumption for every customer. When we get those values, we label them as either "typical", "atypical", "shortterm typical" (this is our solution to the question. I know I wrote "temporarily untypical" earlier, that was wrong, sorry).
When we predict future behaviour, we usually only use the "typical" values from the past. The "atypical" values are considered correct, but not a good information to predict future behaviour. The "shortterm typical" values are for special times. For example when a customer is in an economic crisis, his yesterday's product consumption will be good information to predict his product consumption for tomorrow; but it is not good information to predict his product consumption in 2-3 years.