Altair RISE

A program to recognize and reward our most engaged community members

Nominate Yourself Now!

"C parser for decision trees"

Hi,

I'm generating decision trees which are constructed from a CSV file. The content
of the file (examples for my training set) looks as follows:

a b c label
1 3 5 0
3 5 2 1
...

The decision tree represents a classification that is used to steer
a particular routine within a C++ program. The column notations (a,b,c)
are real variables within the routine. Up to now, I had to translate
the decision tree by hand into equivalent if-then-else C++ statements.
This is very tedious for larger trees. Moreover, I would like to use the
leave-one-one cross-validation for which I have to generate as many
decision trees as many examples I have. This cannot be done manually
in an acceptable amount of time.

Do you know a way how to automatically translate RapidMiner's decision
trees into C++ code like
if( a < 10 ) {
if( b > 20 )
return true;
else if( c < 100 )
return false;
}

Maybe there is a parser plug-in for that.

Thank you for your help.

Regards,
Tim

Find more posts tagged with

AI Studio

Decision Tree

Accepted answers

All comments

IngoRM

Hi Tim,

I am sorry but I am not aware of any implementation of that. Maybe somebody else?

Cheers,
Ingo

Roger DeAngelis_21630

Generating complete case logic for decision trees using altair personal slc

github
https://github.com/rogerjdeangelis/utl-generating-complete-case-when-logic-for-decision-trees-using-altair-personal-slc

hi res tree graphic
https://github.com/rogerjdeangelis/utl-generating-complete-case-when-logic-for-decision-trees-using-altair-personal-slc

Generating complete case logic for decision trees using Altair Personal slc

Note: this is a very simple decision tree example with no holdout sample, no confusion matrix,
no lift charts, no roc curves, no pruning. no logistic analysis and no discriminant analysis.

TWO SOLUTIONS

1 r case when code (generate dynamically - works with most sql dialects))

  case

  when ( score <= 38.31 and age <= 6 and score > 30.77 ) then 'Y'

  when ( score > 38.31 and age <= 8 ) then 'Y'

  when ( score > 38.31 and age > 8 and score > 50.28 ) then 'Y'

  else 'N'

  end as response  




2 python only code (only works in python, Also very different from other languages. forced indent)

  if score <= 38.61000:

      if age <= 6.50000:

          if score <= 31.23500:

              return 0

          else:

              return 1

      else:

          return 0

  else:

      if age <= 8.50000:

          return 1

      else:

          if score <= 50.54500:

              return 0

          else:

       return 1

NOTES:

sas proc hpsplt can produce if/then code. but not sql code.
Unfortunately proc hpsplit is not supported in the Altair Personal SLC
499 proc hpsplit
^
ERROR: Procedure HPSPLIT not known
I believe Altair Rapid Miner can produce if/then else code.

4, Generally you build a model on a random sample, say 100,000 obs then
you apply the model to a much larger population, 100,000,000.
Develop on a power workstation.

5 many decision tree applications allow you to run the code within the application
on larger populations, but having the sql code allows, excel,
spss, matlab any language with an ODBC to apply the model.