What inspired this wish list request?
JMP has great visualization and pre-processing capabilities.
It will be nice to utilize these capabilities for the purpose of generating an explicit expression that is more versatile than polynomial approximation.
Some fields require explicit approximations rather than approximations generated via Machine Learning ("black-box"-type). Quite often pure polynomial approximations are not enough. Symbolic Regression can greatly help.
No initial specification of functional form is required.
https://en.wikipedia.org/wiki/Symbolic_regression
What is the improvement you would like to see?
To have capabilities for Symbolic regression as a part of the Predictive Capabilities.
Why is this idea important?
Symbolic Regression significantly expands the polynomial regression capabilities.
It approaches the versatility of Neural Networks with advantage of providing explicit functional expression as a result.
No initial specification of functional form is required.