I would suggest checking out "Programming Collective Intelligence" and review their chapter on decision trees. That way you would be able to build a 'model' of

I would suggest checking out "Programming Collective Intelligence" and review their chapter on decision trees. That way you would be able to build a 'model' of 'trends' based on "reads" per "time quotient" and use that for your classifier.