Aggregation of Discounts



I am trying to predict sales quantity of an item based on their attributes. Discount is one of those attributes. The problem is I am having different discounts in same period for same item .I need to consider sales for period wise i.e week .While aggregating I am facing problem

on different days item has different discounts.
Case1: Some items are having large variations in discounts like(0% and 50%)
Case2: some are having small variation like(40% and 49%).

-In case1 If we aggregate the discount and divide by sales quantity then there will be decrease of discount in one record and increase in another.

  • In case2 If we aggregate discounts and take average of that .It seems there is no large variations in discount by taking average.Which is not correct for first case.

In my data Sales is more sensitive to discounts.

Is there any way to get generalized discount to predict sales(with Random forest algorithm)?