Predictor Class

The predictor class is used for customizing how tree.predict functions. The defaults can be seen below.

Predictor

Predictor(X, Y, root)

The base Predictor class from which all other predict classes need to inhert.

forest_predict staticmethod

forest_predict(X_old, Y_old, X_new, trees, parallel, **kwargs)

Internal function used by the RandomForest class 'predict' method to evaluate predictions for each tree and aggregate them.

Needs to be adjusted whenever RandomForest predictions do not simply aggregate the tree predictions by averaging.

Parameters:

Name	Type	Description	Default
X_old	ndarray	Array of feature values used during training.	required
Y_old	ndarray	Array of response values used during training.	required
X_new	ndarray	Array of new feature values at which to predict.	required
trees	list[DecisionTree]	List of fitted DecisionTrees fitted within the random forest.	required
parallel	ParallelModel	ParallelModel used for multiprocessing.	required

Returns:

Type	Description
ndarray	An array with predictions for each row of X_new.

predict

predict(X, **kwargs)

Prediction function called by the DecisionTree when it is told to predict. Can be customised for a different output of the DecisionTree.

Parameters:

Name	Type	Description	Default
X	ndarray	A 2-dimensional array for which the rows are the samples at which to predict.	required

Returns:

Type	Description
ndarray	An array with predictions for each row of X.

predict_leaf

predict_leaf(X)

Computes hash table indexing in which LeafNodes the rows of X fall into.

Parameters:

Name	Type	Description	Default
X	ndarray	2-dimensional array for which the rows are the samples at which to predict.	required

Returns:

Type	Description
dict	A hash table with keys corresponding to LeafNode ids and values lists of indices specifying which rows land in a given LeafNode.

PredictorClassification

PredictorClassification(X, Y, root)

The default prediction class for the 'Classification' tree type.

predict

predict(X, **kwargs)

For each row in X, the method first predicts the LeafNode into which the row falls and then computes the class with the highest number of occurances in that LeafNode.

If the keyword argument 'predict_proba' is set to True. This method outputs class probabilities (i.e., the frequence of each label in the same LeafNode).

Parameters:

Name	Type	Description	Default
X	ndarray	A 2-dimensional array for which the rows are the samples at which to predict.	required
**kwargs		predict_proba : bool Specifies whether to compute class probabilities or not. Defaults to False if not provided.	{}

PredictorRegression

PredictorRegression(X, Y, root)

The default prediction class for the 'Regression' tree type.

predict

predict(X, **kwargs)

For each row in X, the method first predicts the LeafNode into which the row falls and then computes average response value in that LeafNode.

Parameters:

Name	Type	Description	Default
X	ndarray	A 2-dimensional array for which the rows are the samples at which to predict.	required

Returns:

Type	Description
ndarray	An array with predictions for each row of X.

PredictorLocalPolynomial

PredictorLocalPolynomial(X, Y, root)

The default prediction class for the 'Gradient' tree type.

predict

predict(X, **kwargs)

For each row in X, the method first predicts the LeafNode into which the row falls and then computes uses the parameters theta0, theta1 and theta2 saved on the LeafNode to estimate the predicted value $$ \texttt{theta0} + \texttt{theta1} \texttt{X}[i, 0] + \texttt{theta2} \texttt{X}[i, 0]^2. $$

Note: This predict class requires that the decision tree/random forest uses a LocalPolynomialLeafNode and either LeafBuilderPartialLinear or LeafBuilderPartialQuadratic.

Parameters:

Name	Type	Description	Default
X	ndarray	A 2-dimensional array for which the rows are the samples at which to predict.	required
**kwargs		orders : list[int] | int A list of integers or a single integer specifying which orders should be predicted. Orders 0, 1 and 2 correspond to the 0th, 1st and 2nd order derivative. Default is orders=[0, 1, 2] if not provided.	{}

Returns:

Type	Description
ndarray	A 2-dim array with the predicted values for each row of X and columns corresponding to the orders.

PredictorQuantile

PredictorQuantile(X, Y, root)

The default prediction class for the 'Quantile' tree type.

predict

predict(X, **kwargs)

For each row in X, the method first predicts the LeafNode into which the row falls and then computes the quantiles in that leaf node.

For random forests the quantiles across all trees are computed jointly.

Parameters:

Name	Type	Description	Default
X	ndarray	A 2-dimensional array for which the rows are the samples at which to predict.	required
**kwargs		quantile : list[float] | float A list of quantiles or a single quantile which should be predicted.	{}

Returns:

Type	Description
ndarray	A 1- or 2-dim array with the predicted quantiles for each row of X and columns corresponding to different quantiles (if quantiles was a list).