Interpretable Machine Learning

Decision Tree Learning

Decision trees are one of the leading forms of interpretable machine learning models. They describe how a record’s target feature value (e.g. edibility of a mushroom) depends on the other record’s feature values (e.g. odor, color).

Decision trees:

  • solve any kind of problem at hand (classification or regression)
  • work with numeric and categorical values
  • work with missing values
  • are suitable for high stakes decisions
  • map nonlinear relationships quite well

We use two types of decision tree learning approaches

  • Heuristics-based (iterative tree splitting and pruning process to find tree topology)
  • Optimization-based (single process to find tree topology)

Example mushroom dataset with 8124 instances:

Mushroom Decision Tree

Our Decision Tree Learning Experts

Marcello Novaes

Fernando Lopez

Axel Reichwein

Talk To Us