Hi all,
I'm using a neural net to predict whether or not a customer will prefer to buy retail items at a markdown or if they are likely to buy full-price (i.e., "bargain buyers" vs. traditional shoppers). I'd like to incoporate the opportunity cost of a false positive (giving discounts to what would otherwise be a full-price shopper) using a profit matrix, but changing values on the matrix in the response column properties does not seem to change the confusion matrix in my model output. I can calculate the new confusion matrix by hand by publishing my prediction formula and using the predictions that generate the most profit, but this is inefficient and time-consuming. I'd ideally like to see it as part of the model output. Does anyone know if this is possible for neural nets in JMP?
