A Python package for fitting Quinlan's Cubist regression model.
Inspired by the R wrapper for cubist: https://github.com/topepo/Cubist
Taking inspiration from the R wrapper this is what needs to be done.
- Get cubist compiled on own machine so we have binary available.
- Understand the inputs the binary require. The training dataset needs to conform to an expected format. May need to write python code that converts Pandas dataframe to this.
- Write python code that forks off cubist process with correct arguments and files.
- Write interpreter that translates cubists model definition to executable Python code.
- Write pypi package that bundles all this.
- Enhance package to compile cubist on users machine.
- Make python translation performant by using scipy or numpy.
- Adapt api to conform to scikit-learn.
- Submit package to scikit-learn.