Neural Network (Dynamic Time Series) - get dependencies from inputparameter to outputvalue

Dear all,

I am working on a neural network to predict temperature of a room by measuring 7 physical parameters which are the inputs of the network. Is there a good way (best practice) to evaluate the dependencies how much influence has a input parameter to the output value - e.g. if the input parameter will change, how big is the influence on the output value.

The use case would be, if a could remove some input parameter ( do optimization on the amount of inputs ).

Best regards Dieter

Answer 1

[Greg Heath]

A quick approximate way is to use STEPWISE and/or STEPWISEFIT on a

1. linear model
2. quadratic model

Hope this helps.

Thank you for formally accepting my answer

Greg