Fig. 3.

ANN versus Standard Least Squares Model Comparison. The models were compared using the model comparison dialog in JMP. This table shows the measure of fit of predicted doublings versus the actual doublings using the 24 runs from the historical dataset. Shown are the 7 neural nets with the highest Measure of Fit for Doublings R² of the 32 created, the 91-Gaussian neural net, and the standard least squares model. Median of root average squared error (RASE) and average absolute error (AAE) for each model are also shown. Figure 3 was created in JMP Pro 14