Neural nets (and Neural-Tools) can also be used when the dependent variable is continuous, not categorical. In this case, the method is an alternative to multiple regression. The file P17_17.xlsx contains fairly old Boston housing data that appears frequently in data mining discussions. The original purpose was to use regression to see how the median value of a house (MEDVAL) depends on air quality, as measured by the variable NOX, after controlling for a number of other variables. The results of this regression appear in the Original Analysis sheet. As you can see on this sheet, the resulting RMSE (root mean squared error) is about 4 (in thousands of dollars). See if you can get better results with Neural-Tools. Unlike the regression analysis, you don’t need to create any nonlinear transformations—Neural-Tools does this internally. Use the GRN algorithm, make sure the Perform Linear Regression option in the Net Configuration tab is unchecked, and don’t use any test cases. That is, use all of the cases for training. Discuss your results?