Tree API

Classes / functions for creating and modifying decision trees. More...

Classes
class	StochTree::Tree
	Decision tree data structure. More...
class	StochTree::TreeSplit
	Representation of arbitrary tree split rules, including numeric split rules (X[,i] <= c) and categorical split rules (X[,i] in {2,4,6,7}) More...

Functions
bool	StochTree::operator== (const Tree &lhs, const Tree &rhs)
	Comparison operator for trees.
bool	StochTree::SplitTrueNumeric (double fvalue, double threshold)
	Determine whether an observation produces a "true" value in a numeric split node.
bool	StochTree::SplitTrueCategorical (double fvalue, std::vector< std::uint32_t > const &category_list)
	Determine whether an observation produces a "true" value in a categorical split node.
int	StochTree::NextNodeNumeric (double fvalue, double threshold, int left_child, int right_child)
	Return left or right node id based on a numeric split.
int	StochTree::NextNodeCategorical (double fvalue, std::vector< std::uint32_t > const &category_list, int left_child, int right_child)
	Return left or right node id based on a categorical split.
int	StochTree::EvaluateTree (Tree const &tree, Eigen::MatrixXd &data, int row)
int	StochTree::EvaluateTree (Tree const &tree, Eigen::Map< Eigen::Matrix< double, Eigen::Dynamic, Eigen::Dynamic, Eigen::ColMajor > > &data, int row)
bool	StochTree::RowSplitLeft (Eigen::MatrixXd &covariates, int row, int split_index, double split_value)
	Determine whether a given observation is "true" at a split proposed by split_index and split_value.
bool	StochTree::RowSplitLeft (Eigen::MatrixXd &covariates, int row, int split_index, std::vector< std::uint32_t > const &category_list)
	Determine whether a given observation is "true" at a split proposed by split_index and split_value.

Detailed Description

Classes / functions for creating and modifying decision trees.

Design

Function Documentation

SplitTrueNumeric()

bool StochTree::SplitTrueNumeric ( double  fvalue, double  threshold  )

inline

Determine whether an observation produces a "true" value in a numeric split node.

Parameters

fvalue	Value of the split feature for the observation
threshold	Value of the numeric split threshold at the node
left_child	Node id of the left child
right_child	Node id of the right child

SplitTrueCategorical()

bool StochTree::SplitTrueCategorical ( double  fvalue, std::vector< std::uint32_t > const &  category_list  )

inline

Determine whether an observation produces a "true" value in a categorical split node.

Parameters

fvalue	Value of the split feature for the observation
category_list	Category indices that route an observation to the left child
left_child	Node id of the left child
right_child	Node id of the right child

NextNodeNumeric()

int StochTree::NextNodeNumeric ( double  fvalue, double  threshold, int  left_child, int  right_child  )

inline

Return left or right node id based on a numeric split.

Parameters

fvalue	Value of the split feature for the observation
threshold	Value of the numeric split threshold at the node
left_child	Node id of the left child
right_child	Node id of the right child

NextNodeCategorical()

int StochTree::NextNodeCategorical ( double  fvalue, std::vector< std::uint32_t > const &  category_list, int  left_child, int  right_child  )

inline

Return left or right node id based on a categorical split.

Parameters

fvalue	Value of the split feature for the observation
category_list	Category indices that route an observation to the left child
left_child	Node id of the left child
right_child	Node id of the right child

EvaluateTree() [1/2]

int StochTree::EvaluateTree ( Tree const &  tree, Eigen::MatrixXd &  data, int  row  )

inline

Determine the node at which a tree places a given observation

Parameters

tree	Tree object used for prediction
data	Dataset used for prediction
row	Row indexing the prediction observation

EvaluateTree() [2/2]

int StochTree::EvaluateTree ( Tree const &  tree, Eigen::Map< Eigen::Matrix< double, Eigen::Dynamic, Eigen::Dynamic, Eigen::ColMajor > > &  data, int  row  )

inline

Determine the node at which a tree places a given observation

Parameters

tree	Tree object used for prediction
data	Dataset used for prediction
row	Row indexing the prediction observation

RowSplitLeft() [1/2]

bool StochTree::RowSplitLeft ( Eigen::MatrixXd &  covariates, int  row, int  split_index, double  split_value  )

inline

Determine whether a given observation is "true" at a split proposed by split_index and split_value.

Parameters

covariates	Dataset used for prediction
row	Row indexing the prediction observation
split_index	Column of new split
split_value	Value defining the split

RowSplitLeft() [2/2]

bool StochTree::RowSplitLeft ( Eigen::MatrixXd &  covariates, int  row, int  split_index, std::vector< std::uint32_t > const &  category_list  )

inline

Determine whether a given observation is "true" at a split proposed by split_index and split_value.

Parameters

covariates	Dataset used for prediction
row	Row indexing the prediction observation
split_index	Column of new split
category_list	Categories defining the split