convert variables names to original

[jermaine1ronquillo]

Is it possible to convert the variables in the equation to its original name (instead of X1, X2...), and add mapping function to shorten variable names in case the names are too long? Can we use sympy to convert the equation? Thank you

[trevorstephens]

Where would you want to see this representation? The print function, or the program tree plot, or both? The print function could be problematic due to the feature names being related to the data and hence should only be passed to fit if conforming to the scikit-learn standard API. The graphviz plot might be more of an option as this is its own method and happens after fitting.

As for sympy I don't particularly want to go down that road as many of the built-in functions rely on closure which means there are a lot of rounding checks that don't fit neatly into the sympy world and user-defined functions add a whole other level of complexity to that. Happy to hear further thoughts on this or a PR but it is not something that I am prioritising right now.

[jermaine1ronquillo]

Thank you for the response, the graphviz plot will do.

[trevorstephens]

Hey @jermaine1ronquillo ... I know it's been a while but would proposed changes in #119 work for you API wise? Allows specifying feature names during init of estimator for both print and graphviz, but I might think about how I fail here. Might be better to warn and output X's than error and give nothing.