Welcome to monoensemble’s documentation!

This package contains two key classification algorithms:

• MonoRandomForestClassifier is a Random Forest classifier with the added capability of partially monotone features. It is very fast and demonstrates excellent experimental accuracy.
• MonoGradientBoostingClassifier is a monotone Gradient Boosting classifier. It is very fast and demonstrates very good experimental accuracy.

These algorithms are heavily based on (or inherit from) sci-kit learn’s versions, and the interface is identical except that the constructor has three additional parameters:

• incr_feats : The one-based array indices of the columns in X that should only have a monotone increasing impact on the resulting class.
• decr_feats : The one-based array indices of the columns in X that should only have a monotone decreasing impact on the resulting class.

• coef_calc_type : string

  Determines how the rule coefficients are calculated. Allowable values:

  • ‘boost’ DEFAULT: A single Newton step approximation is used. Fast, and generally best.
  • ‘bayesian’: Assumes conditional independence between rules and calculates coefficients as per Naive bayesian classification. Fast with good results.
  • ‘logistic’: L2 regularised logistic regression. Slower.

To install, simply use pip install monoensemble or conda install -c chriswbartley monoensemble. For full documentation you’ve come to the right place. For a brief overview, refer to the README file in the Github repository.

Contents:

• Theory
• MonoRandomForestClassifier Examples
  • Basic Model Fitting - MonoGradientBoostingClassifier
  • Basic Model Fitting - MonoRandomForestClassifier
• API
  • Classes
  • Functions