The apriori file gives a weight to the input classes that will be used during the classification. Each cell in the classification has a probability of being assigned to each class. The cell will be assigned to the class that it has the highest probability of being. In certain situations, you know that a cell is more likely to be part of one class over the other.
For example, if you are classifying a coastal area, it is more likely that a cell will be water than another land use. Since this apriori information defines a probability, to normalize the probabilities, they should add up to 1.