Table of Contents

Interface ITreeBasedRegression<T>

Namespace
AiDotNet.Interfaces
Assembly
AiDotNet.dll

Defines the core functionality for tree-based machine learning models.

public interface ITreeBasedRegression<T> : INonLinearRegression<T>, IRegression<T>, IFullModel<T, Matrix<T>, Vector<T>>, IModel<Matrix<T>, Vector<T>, ModelMetadata<T>>, IModelSerializer, ICheckpointableModel, IParameterizable<T, Matrix<T>, Vector<T>>, IFeatureAware, IFeatureImportance<T>, ICloneable<IFullModel<T, Matrix<T>, Vector<T>>>, IGradientComputable<T, Matrix<T>, Vector<T>>, IJitCompilable<T>

Type Parameters

T

The numeric data type used for calculations (e.g., float, double).

Inherited Members
Extension Methods

Remarks

Tree-based models make predictions by following a series of decision rules organized in a tree-like structure. These models can be used for both classification (predicting categories) and regression (predicting numeric values).

For Beginners: Tree-based models work like a flowchart of yes/no questions to make predictions. Imagine you're trying to predict if someone will like a movie:

  1. Is it an action movie? If yes, go to question 2. If no, go to question 3.
  2. Does it have their favorite actor? If yes, predict "Like". If no, predict "Dislike".
  3. Is it less than 2 hours long? If yes, predict "Like". If no, predict "Dislike".

This is a simple decision tree. More advanced tree-based models like Random Forests or Gradient Boosted Trees use multiple trees together to make better predictions.

This interface inherits from IFullModel<T>, which provides the basic methods for training, predicting, and evaluating machine learning models.

Properties

FeatureImportances

Gets the relative importance of each feature in making predictions.

Vector<T> FeatureImportances { get; }

Property Value

Vector<T>

Remarks

Feature importance indicates how much each input variable contributes to the model's predictions. Higher values indicate more important features.

For Beginners: This tells you which of your input variables are most helpful for making predictions.

For example, if you're predicting house prices:

  • FeatureImportances[0] = 0.7 for "square footage"
  • FeatureImportances[1] = 0.2 for "number of bedrooms"
  • FeatureImportances[2] = 0.1 for "year built"

This would tell you that square footage is the most important factor in your model's predictions, followed by number of bedrooms, with year built having the least impact.

You can use this information to:

  • Focus on collecting better data for important features
  • Possibly remove unimportant features to simplify your model
  • Better understand what drives the predictions in your specific problem

MaxDepth

Gets the maximum depth (number of sequential decisions) allowed in each decision tree.

int MaxDepth { get; }

Property Value

int

Remarks

The depth of a tree is the maximum number of decisions that must be made to reach a prediction.

For Beginners: This tells you how many questions the model can ask before making a prediction.

For example, if MaxDepth = 3:

  • The model can ask at most 3 questions before making a prediction
  • This creates a simpler model that might be easier to understand
  • But it might miss complex patterns in your data

If MaxDepth = 20:

  • The model can ask up to 20 questions before deciding
  • This creates a more complex model that can capture detailed patterns
  • But it might "memorize" your training data instead of learning general rules

Setting the right MaxDepth helps balance between a model that's too simple (underfitting) and one that's too complex (overfitting).

NumberOfTrees

Gets the number of decision trees used in the model.

int NumberOfTrees { get; }

Property Value

int

Remarks

For single decision tree models, this value is 1. For ensemble methods like Random Forests or Gradient Boosted Trees, this represents the number of trees in the ensemble.

For Beginners: Think of this as "how many different flowcharts is the model using to make its decision?"

  • If NumberOfTrees = 1: The model is using a single decision tree (like the movie example above)
  • If NumberOfTrees > 1: The model is using multiple trees and combining their predictions

More trees often lead to better predictions but make the model slower and more complex. Common values range from 10 to 1000 trees, depending on the specific algorithm and dataset.

Edit this page