tree

Type Members

1. final case class CategoricalSplit(featureIndex: Int, numCategories: Int, categories: Array[Double], isLeft: Boolean) extends Split with Product with Serializable

   

   Class for splitting on a categorical feature.

   Class for splitting on a categorical feature.

   featureIndex

   index of the features

   numCategories

   number of potential categories

   categories

   array of categories

   isLeft

   if this split goes left

   Annotations

   @SparkCode()

2. final case class ContinuousSplit(featureIndex: Int, threshold: Double) extends Split with Product with Serializable

   

   Class for splitting on a continuous feature.

   Class for splitting on a continuous feature.

   featureIndex

   index of the features

   threshold

   threshold for going left or right

   Annotations

   @SparkCode()

3. trait DecisionTree extends Serializable

   

   Trait for a decision tree.

4. final case class InternalNode(left: Node, right: Node, split: Split) extends Node with Product with Serializable

   

   Trait for internal node in a decision tree.

   Trait for internal node in a decision tree.

   left

   left child

   right

   right child

   split

   split logic to go left or right

   Annotations

   @SparkCode()

5. final case class LeafNode(prediction: Double, impurities: Option[Vector] = None) extends Node with Product with Serializable

   

   Trait for a leaf node in a decision tree.

   Trait for a leaf node in a decision tree.

   prediction

   prediction for this leaf node

   impurities

   options vector of impurities

   Annotations

   @SparkCode()

6. sealed trait Node extends Serializable

   

   Trait for a node in a decision tree.

   Trait for a node in a decision tree.

   Annotations

   @SparkCode()

7. sealed trait Split extends Serializable

   

   Trait for a split logic.

   Trait for a split logic.

   Normally used for splits in a decision tree.

   Annotations

   @SparkCode()

8. trait TreeEnsemble extends AnyRef

   

   Trait for an ensemble of decision trees.