I have used a dataset prepared from forest fire inventory and factor rasters to train an ANN to predict forest fire of an area. The ANN performance is satisfactory.

I have with me the weights/ importance values of the factors (independent variables) and parameters of the ANN.

The activation function for the hidden layer is Hyperbolic tangent and that of the output layer is softmax.

How do I prepare the prediction map from these data? Should I use a weighted linear combination method of the factor rasters and factor weights to prepare the prediction map?

Is this the correct method?