Tree Models

Tree models predict values by recursively partitioning data into subsets based on feature thresholds, creating a structure that resembles an upside-down tree. Each internal node represents a decision point, while the terminal “leaves” provide the final numerical prediction or category. Their main appeal lies in their interpretability and their ability to capture complex, non-linear relationships without requiring extensive data preprocessing.

Pros tree models:

  • Can understand non-linear and complex relationships.

Cons tree models:

  • Very Prone to over fitting

Regressors

Model NameSummaryLink to sklearn-documentation
DecisionTreeRegressorA non-parametric supervised learning method that predicts a continuous target value by learning simple decision rules (if-then-else) inferred from the data features.Documentation
ExtraTreeRegressorAn extremely randomized tree regressor that introduces additional randomness by choosing a split point at random for each feature instead of searching for the optimal threshold, helping to reduce variance.Documentation

Classifiers

Model NameSummaryLink to sklearn-documentation
**DecisionTreeClassifierA non-parametric supervised learning method that predicts a continuous target value by learning simple decision rules (if-then-else) inferred from the data features.Documentation
ExtraTreeClassifierAn extremely randomized tree regressor that introduces additional randomness by choosing a split point at random for each feature instead of searching for the optimal threshold, helping to reduce variance.Documentation