LightGBM::Tree Class Reference

Tree model. More...

#include <tree.h>

Public Member Functions
	Tree (int max_leaves)
	Constructor.
	Tree (const char *str, size_t *used_len)
	Construtor, from a string.
int	Split (int leaf, int feature, int real_feature, uint32_t threshold_bin, double threshold_double, double left_value, double right_value, int left_cnt, int right_cnt, float gain, MissingType missing_type, bool default_left)
	Performing a split on tree leaves.
int	SplitCategorical (int leaf, int feature, int real_feature, const uint32_t *threshold_bin, int num_threshold_bin, const uint32_t *threshold, int num_threshold, double left_value, double right_value, int left_cnt, int right_cnt, float gain, MissingType missing_type)
	Performing a split on tree leaves, with categorical feature.
double	LeafOutput (int leaf) const
	Get the output of one leaf.
void	SetLeafOutput (int leaf, double output)
	Set the output of one leaf.
void	AddPredictionToScore (const Dataset *data, data_size_t num_data, double *score) const
	Adding prediction value of this tree model to scores.
void	AddPredictionToScore (const Dataset *data, const data_size_t *used_data_indices, data_size_t num_data, double *score) const
	Adding prediction value of this tree model to scorese.
double	Predict (const double *feature_values) const
	Prediction on one record.
double	PredictByMap (const std::unordered_map< int, double > &feature_values) const
int	PredictLeafIndex (const double *feature_values) const
int	PredictLeafIndexByMap (const std::unordered_map< int, double > &feature_values) const
void	PredictContrib (const double *feature_values, int num_features, double *output)
int	num_leaves () const
	Get Number of leaves.
int	leaf_depth (int leaf_idx) const
	Get depth of specific leaf.
int	split_feature (int split_idx) const
	Get feature of specific split.
double	split_gain (int split_idx) const
int	data_count (int node) const
	Get the number of data points that fall at or below this node.
void	Shrinkage (double rate)
	Shrinkage for the tree's output shrinkage rate (a.k.a learning rate) is used to tune the traning process.
double	shrinkage () const
void	AddBias (double val)
void	AsConstantTree (double val)
std::string	ToString () const
	Serialize this object to string.
std::string	ToJSON () const
	Serialize this object to json.
std::string	ToIfElse (int index, bool predict_leaf_index) const
	Serialize this object to if-else statement.
void	RecomputeMaxDepth ()

Static Public Member Functions
static bool	IsZero (double fval)
static bool	GetDecisionType (int8_t decision_type, int8_t mask)
static void	SetDecisionType (int8_t *decision_type, bool input, int8_t mask)
static int8_t	GetMissingType (int8_t decision_type)
static void	SetMissingType (int8_t *decision_type, int8_t input)

Detailed Description

Tree model.

Constructor & Destructor Documentation

Tree() [1/2]

LightGBM::Tree::Tree ( int  max_leaves )

explicit

Constructor.

Parameters

max_leaves

The number of max leaves

Tree() [2/2]

LightGBM::Tree::Tree	(	const char *	str,
		size_t *	used_len
	)

Construtor, from a string.

Parameters

str	Model string
used_len	used count of str

Member Function Documentation

AddPredictionToScore() [1/2]

void LightGBM::Tree::AddPredictionToScore	(	const Dataset *	data,
		const data_size_t *	used_data_indices,
		data_size_t	num_data,
		double *	score
	)		const

Adding prediction value of this tree model to scorese.

Parameters

data	The dataset
used_data_indices	Indices of used data
num_data	Number of total data
score	Will add prediction to score

AddPredictionToScore() [2/2]

void LightGBM::Tree::AddPredictionToScore	(	const Dataset *	data,
		data_size_t	num_data,
		double *	score
	)		const

Adding prediction value of this tree model to scores.

Parameters

data	The dataset
num_data	Number of total data
score	Will add prediction to score

Predict()

double LightGBM::Tree::Predict ( const double *  feature_values ) const

inline

Prediction on one record.

Parameters

feature_values

Feature value of this record

Returns

Prediction result

Shrinkage()

void LightGBM::Tree::Shrinkage ( double  rate )

inline

Shrinkage for the tree's output shrinkage rate (a.k.a learning rate) is used to tune the traning process.

Parameters

rate	The factor of shrinkage

Split()

int LightGBM::Tree::Split	(	int	leaf,
		int	feature,
		int	real_feature,
		uint32_t	threshold_bin,
		double	threshold_double,
		double	left_value,
		double	right_value,
		int	left_cnt,
		int	right_cnt,
		float	gain,
		MissingType	missing_type,
		bool	default_left
	)

Performing a split on tree leaves.

Parameters

leaf	Index of leaf to be split
feature	Index of feature; the converted index after removing useless features
real_feature	Index of feature, the original index on data
threshold_bin	Threshold(bin) of split
threshold_double	Threshold on feature value
left_value	Model Left child output
right_value	Model Right child output
left_cnt	Count of left child
right_cnt	Count of right child
gain	Split gain
missing_type	missing type
default_left	default direction for missing value

Returns

The index of new leaf.

SplitCategorical()

int LightGBM::Tree::SplitCategorical	(	int	leaf,
		int	feature,
		int	real_feature,
		const uint32_t *	threshold_bin,
		int	num_threshold_bin,
		const uint32_t *	threshold,
		int	num_threshold,
		double	left_value,
		double	right_value,
		int	left_cnt,
		int	right_cnt,
		float	gain,
		MissingType	missing_type
	)

Performing a split on tree leaves, with categorical feature.

Parameters

leaf	Index of leaf to be split
feature	Index of feature; the converted index after removing useless features
real_feature	Index of feature, the original index on data
threshold_bin	Threshold(bin) of split, use bitset to represent
num_threshold_bin	size of threshold_bin
threshold	Thresholds of real feature value, use bitset to represent
num_threshold	size of threshold
left_value	Model Left child output
right_value	Model Right child output
left_cnt	Count of left child
right_cnt	Count of right child
gain	Split gain

Returns

The index of new leaf.

---

The documentation for this class was generated from the following files:

• External/LightGBM_2.2.3/LightGBM-2.2.3/include/LightGBM/tree.h
• External/LightGBM_2.2.3/LightGBM-2.2.3/src/io/tree.cpp