Type Members

1.  final case class DecisionTree(root: Node) extends Product with Serializable

   An XGboost Decision Tree Note that algorithms are independent of Nodes implementation as Node is abstract

   An XGboost Decision Tree Note that algorithms are independent of Nodes implementation as Node is abstract

   root

   root node

2.  final case class DenseFVector(values: Array[Double], treatZeroAsNa: Boolean = true) extends FVector with Product with Serializable

   Dense Feature vector

   Dense Feature vector

   values

   features data

   treatZeroAsNa

   considers 0.0 as N/A

3.  trait FVector extends Serializable

   A generic Feature Vector

4.  final case class ModelParams(baseScore: Float, nbFeatures: Int, nbClass: Int, nbTrees: Int, nbGroups: Int, objName: String) extends Product with Serializable

   the model Parameters

   the model Parameters

   baseScore

   base score

   nbFeatures

   number of features

   nbClass

   number of classes

   nbTrees

   number of trees

   nbGroups

   numbet of groups

   objName

   objective name

5.  abstract class Node extends Serializable

   A generic (binary) node of a decision Tree

6.  abstract class NodeFactory extends AnyRef
7.  final case class NodeImpl(left: Node, right: Node, id: Int, index: Int, defaultLeft: Boolean, value: Float, sumHess: Double, mean: Double = 0.0) extends Node with Product with Serializable

   The node implementation (actually a simple case class)

   The node implementation (actually a simple case class)

   left

   left child

   right

   right child

   id

   rw node id (just used for information)

   index

   split index

   defaultLeft

   is left child the default

   value

   node value

   sumHess

   sum hess

   mean

   mean value

8.  trait Objective extends Serializable

   The XGboost Objective functions

9.  abstract class PathElement extends AnyRef

   Element of a unique Path in SHAP algorithm https://arxiv.org/pdf/1802.03888.pdf

10.  final case class PathElementImpl(parent: PathElement, p0: Double, p1: Double, index: Int, depth: Double, weight: Double) extends PathElement with Product with Serializable
11.  case class Predictor(params: ModelParams, trees: Seq[DecisionTree], objective: Objective) extends Product with Serializable

    Predictor fox XGboost

    Predictor fox XGboost

    params

    model parameters

    trees

    decision tree

    objective

    objective function

    Annotations

    @SerialVersionUID()

12.  class PredictorBuilder extends AnyRef

    Builder to ease the creation of Predictor from Java (avoid using Scala Module name)

13.  final case class RawNode(id: Int, left: Int, right: Int, isLeaf: Boolean, value: Float, defaultLeft: Boolean, splitIndex: Int, sumHess: Float) extends Product with Serializable

    RawNode is just use to read the trees from binary model file

    RawNode is just use to read the trees from binary model file

    id

    node id

    left

    left node id

    right

    right node id

    isLeaf

    is this a leaf node

    value

    leaf value or split value

    defaultLeft

    is left child the default

    splitIndex

    split index

    sumHess

    sum Hess

14.  final case class SparseFVector(values: Map[Int, Double]) extends FVector with Product with Serializable

    Sparse Feature vector

    Sparse Feature vector

    values

    a map of position/value