cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Check out the JMP® Marketplace featured Capability Explorer add-in
Choose Language Hide Translation Bar
MD513
MD513
Level I

Neural Net JMP 14 SAS Script - Converting to excel formula

Jan 30, 2019 04:53 PM (2827 views)

Hi All,

 

I am relatively new to JMP and have created a Neural Network model that fits pretty well. I am trying to implement the output (model) in excel. I exported the SAS script and currently I am trying to convert the formula in the script to excel. The following formula in the script is tripping me. 

H1 = tanh(.5*(-1.31215248048067*((ENGINE_MODEL="XXX")-(ENGINE_MODEL="YYY"))   + -0.0627257389428689*......

Engine_Model is a binary input to the model (can be XXX (0) or YYY(1)). 

 

I need soem help in interpreting the term (ENGINE_MODEL="XXX")-(ENGINE_MODEL="YYY"). Does it simply mean that it is difference between the value of ENGINE_MODEL="XXX" and ENGINE_MODEL="YYY"? In that case it is just a constant number. Also the calculated value of H1 in excel and JMP do not match with that assumption.

 

Any help is appreciated.

0 Kudos
1 REPLY 1
Phil_Kay
Staff Phil_Kay
Staff

Re: Neural Net JMP 14 SAS Script - Converting to excel formula

Created: Feb 18, 2019 02:53 AM  |  Last Modified: Feb 17, 2019 11:54 PM (2725 views)  |  Posted in reply to message from MD513 01-30-2019

Hi,

I think this is about how JMP codes categorical factors in regression models (Neural Nets, although exotic, are just complex regression models).

 

This from JMP's statistics knowledge portal should help.

 

"To integrate a two-level categorical variable into a regression model, we create one indicator or dummy variable with two values..."

 

I would suggest you experiment with a very simple Neural Net. For example, a neural with only Engine Model and 1 node. Then check that the JMP prediction and your conversion of this simple model are in agreement.