More on the Classification Script

I am currently reading the Prentice Hall book, "Artificial Intelligence, a Modern Approach." Its stimulating some thoughts on optimization of the Classification script. I am thinking now of instead of classifying each record, it might be beneficial to go through the records to create and modify the classifications, then go through them again to actually classify them. This should eliminate much of the reclassification that takes place throughout much of the run time. Instead of going through all of the records for each new classification, it should allow the proper classification in (hopefully) a single pass. The classifications do not depend upon the members of the classification, only upon there being at least one member. Instead, each class could store the number of potential members, and could be eliminated when that number drops to zero. I'll have to work on it, but I think it may have some potential to significantly reduce the time required.