Decision Forest
Details
Given n feature vectors 
of
-dimensional feature vectors and n responses 
, the problem is to build a decision forest
classification or regression model.
of
-dimensional feature vectors and n responses , the problem is to build a decision forest
classification or regression model.Training Stage
Library uses the following algorithmic framework for the training
stage. Let 
be the set of observations. Given a positive
integer parameters, such as the number of trees 
, the bootstrap
parameter 
, where 
is a fraction of observations used for
a training of one tree, and the number of features per node 
, the
algorithm does the following for 
:
be the set of observations. Given a positive
integer parameters, such as the number of trees , the bootstrap
parameter , where is a fraction of observations used for
a training of one tree, and the number of features per node , the
algorithm does the following for :- Selects randomly with replacement the set
of 
vectors from the set 
. The set 
is called a
bootstrapset. - Trains a decision tree classifier
on
using parameter for each tree.
Decision tree 
is trained using the training set 
of size .
Each node 
in the tree corresponds to the subset 
of
the training set , with the root node being itself. Its
internal nodes represent a binary test (split) dividing their
subset 
in two subsets 
and 
,
corresponding to their children 
and 
.
is trained using the training set of size .
Each node in the tree corresponds to the subset of
the training set , with the root node being itself. Its
internal nodes represent a binary test (split) dividing their
subset in two subsets and ,
corresponding to their children and .Inexact Histogram Computation Method
In inexact histogram method only a selected subset of splits is considered for computation of a best split.
This subset is computed for each feature at the initialization stage of the algorithm. After the set of splits is computed,
each value from initially provided data is substituted with the value of the corresponding bin.
The bins are continuous intervals between the selected splits.
Split Criteria
The metric for measuring the best split is called 
. It generally reflects the homogeneity of responses within
the subset in the node .
For the detailed definition of 
metrics, see the description
of a specific algorithm.
impurity
,
. It generally reflects the homogeneity of responses within
the subset in the node .
For the detailed definition of metrics, see the description
of a specific algorithm.Let the 
be
impurity decrease
in the node be
Termination Criteria
The library supports the following termination criteria of
decision forest training:
- Minimal number of observations in a leaf node
- Node is not processed if
is smaller than the predefined value.
Splits that produce nodes with the number of observations smaller than that value are not allowed. - Minimal number of observations in a split node
- Node is not processed if is smaller than the predefined value.
Splits that produce nodes with the number of observations smaller than that value are not allowed.
- Minimum weighted fraction of the sum total of weights of all the input observations required to be at a leaf node
- Node is not processed if is smaller than the predefined value.
Splits that produce nodes with the number of observations smaller than that value are not allowed.
- Maximal tree depth
- Node is not processed if its depth in the tree reached the predefined value.
- Impurity threshold
- Node is not processed if its impurity is smaller than the predefined threshold.
- Maximal number of leaf nodes
- Grow trees with positive maximal number of leaf nodes in a best-first fashion. Best nodes are defined by relative reduction in impurity. If maximal number of leaf nodes equals zero, then this criterion does not limit the number of leaf nodes, and trees grow in a depth-first fashion.
Tree Building Strategies
Maximal number of leaf nodes defines the strategy of tree building:
depth-first or best-first.
Depth-first Strategy
If maximal number of leaf nodes equals zero, a decision tree is built using depth-first strategy.
In each terminal node , the following recursive procedure is applied:
, the following recursive procedure is applied:- Stop if the termination criteria are met.
- Choose randomly without replacement
feature indices 
. - For each
, find the best split 
that
partitions subset and maximizes impurity decrease
. - A node is a split if this split induces a decrease of the impurity greater than or equal to the predefined value. Get the best split
that maximizes impurity decrease
in all splits. - Apply this procedure recursively to
and .
Best-first Strategy
If maximal number of leaf nodes is positive, a decision tree is built using best-first strategy.
In each terminal node , the following steps are applied:
, the following steps are applied:- Stop if the termination criteria are met.
- Choose randomly without replacement feature indices .
- For each, find the best split that
partitions subset and maximizes impurity decrease
.
- A node is a split if this split induces a decrease of the impurity greater than or equal to the predefined value and the number of split nodes is less or equal to
.
Get the best split 
that maximizes impurity decrease
in all splits. - Put a node into a sorted array, where sort criterion is the improvement in impurity
.
The node with maximal improvement is the first in the array. For a leaf node, the improvement in impurity is zero. - Apply this procedure to and and grow a tree one by one getting the first element from the array
until the array is empty.
Random Numbers Generation
To create a
bootstrap
set and choose feature indices in the
performant way, the training algorithm requires the source of
random numbers, capable to produce sequences of random numbers in
parallel.Initialization of the engine in the decision forest is based on
the scheme below:
The state of the engine is updated once the training of the
decision forest model is completed. The library provides support
to retrieve the instance of the engine with updated state that can
be used in other computations. The update of the state is
engine-specific and depends on the parallelization technique used
as defined earlier:
- Family: the updated state is the set of states that represent individual engines in the family.
- Leapfrog: the updated state is the state of the sequence with the rightmost position on the sequence. The example below demonstrates the idea for case of 2 subsequences (‘x’ and ‘o’) of the random number sequence:
- SkipAhead: the updated state is the state of the independent sequence with the rightmost position on the sequence. The example below demonstrates the idea for case of 2 subsequences (‘x’ and ‘o’) of the random number sequence:
Prediction Stage
Given decision forest classifier and vectors 
,
the problem is to calculate the responses for those vectors. To
solve the problem for each given query vector 
, the
algorithm finds the leaf node in a tree in the forest that gives
the response by that tree. The response of the forest is based on
an aggregation of responses from all trees in the forest. For the
detailed definition, see the description of a specific algorithm.
,
the problem is to calculate the responses for those vectors. To
solve the problem for each given query vector , the
algorithm finds the leaf node in a tree in the forest that gives
the response by that tree. The response of the forest is based on
an aggregation of responses from all trees in the forest. For the
detailed definition, see the description of a specific algorithm.Additional Characteristics Calculated by the Decision Forest
Decision forests can produce additional characteristics, such as
an estimate of generalization error and an importance measure
(relative decisive power) of each of p features (variables).
Out-of-bag Error
The estimate of the generalization error based on the training
data can be obtained and calculated as follows:
- For each tree in the forest, trained on the bootstrap set , the set
is called the out-of-bag (OOB) set. - Predict the data from
set by 
. - For each vector
in the dataset X, predict its response 
by the trees that contain in their OOB set. - Aggregate the out-of-bag predictions in all trees and calculate the OOB error of the decision forest.
- If OOB error value per each observation is required, then calculate the prediction error for.
For the detailed definition, see the description of a specific
algorithm.
Variable Importance
There are two main types of variable importance measures:
- Mean Decrease Impurityimportance (MDI).Importance of the
-th variable for predicting 
is the sum of
weighted impurity decreases 
for all nodes
that use 
, averaged over all trees in the
forest:
where
is the fraction of observations reaching node
in the tree , and 
is the index of the
variable used in split . - Mean Decrease Accuracy(MDA).Importance of the-th variable for predicting is the average
increase in the OOB error over all trees in the forest when the
values of the
-th variable are randomly permuted in the OOB
set. For that reason, this latter measure is also known as
permutation importance.In more details, the library calculates MDA importance as follows:- Let
be the set of feature vectors where the -th variable is randomly permuted over all vectors in the set. - Let
be the OOB error calculated for 
on its out-of-bag dataset 
. - Let
be the OOB error calculated for using 
, and its out-of-bag dataset is permuted on the -th variable. Then
is the OOB error increase for the tree .
is MDA importance.
, where 
is the variance of 
Batch Processing
Decision forest classification and regression follows the general
workflow described in Classification Usage Model.
Training
At the training stage, decision forest regression has the following parameters:
Parameter | Default Value | Description |
|---|---|---|
nTrees |  | The number of trees in the forest. |
observationsPerTreeFraction |  | Fraction of the training set S used to form the bootstrap set for a
single tree training,  . The
observations are sampled randomly with replacement. |
featuresPerNode |  | The number of features tried as possible splits per node. If the
parameter is set to , the library uses the square root of the number of
features,  , for classification
and  features for regression. |
maxTreeDepth | | Maximal tree depth. Default is (unlimited). |
DEPRECATED: seed |  | The seed for random number generator, which is used to choose the
bootstrap set, split features in every split node in a tree, and
generate permutation required in computations of MDA variable importance.This parameter is deprecated and will be removed in future releases. Use engine instead. |
engine | SharePtr< engines:: mt2203:: Batch>() | Pointer to the random number generator engine. The random numbers produced by this engine are used to choose the bootstrap set,
split features in every split node in a tree, and generate permutation required in computations
of MDA variable importance. |
impurityThreshold | | The threshold value used as stopping criteria: if the impurity value in
the node is smaller than the threshold, the node is not split anymore. |
varImportance | none | The variable importance computation mode. Possible values:
|
resultsToCompute | | The 64-bit integer flag that specifies which extra characteristics of
the decision forest to compute. Provide one of the following values to
request a single characteristic or use bitwise OR to request a
combination of the characteristics:
|
bootstrap | true | If true, the training set for a tree is a bootstrap of the whole training set.
If false, the whole training set is used to build trees. |
minObservationsInLeafNode | for classification,  for regression | Minimum number of observations in the leaf node. |
minObservationsInSplitNode |  | Minimum number of samples required to split an internal node; it can be any non-negative number. |
minWeightFractionInLeafNode |  | Minimum weighted fraction of the sum total of weights of all the input observations required to be at a leaf node,
from to  .All observations have equal weights if the weights of the observations are not provided. |
minImpurityDecreaseInSplitNode | | Minimum amount of impurity decrease required to split a node; it can be any non-negative number. |
maxLeafNodes | | Grow trees with positive maximal number of leaf nodes in a best-first fashion.
Best nodes are defined as relative reduction in impurity.
If maximal number of leaf nodes equals zero,
then this parameter does not limit the number of leaf nodes, and trees grow in a depth-first fashion. |
Output
In addition to regression or classifier output, decision forest
calculates the result described below. Pass the
Result ID
as a
parameter to the methods that access the result of your algorithm.Result ID | Result |
|---|---|
outOfBagError | A numeric table  containing out-of-bag error computed when the
computeOutOfBagErroroption option is on.By default, this result is an object of the HomogenNumericTable class,
but you can define the result as an object of any class derived from NumericTable . |
variableImportance | A numeric table  that contains variable importance values for each
feature. If you set the varImportance parameter to none, the library
returns a null pointer to the table.By default, this result is an object of the HomogenNumericTable class,
but you can define the result as an object of any class derived from NumericTable
except PackedTriangularMatrix and PackedSymmetricMatrix . |
outOfBagErrorPerObservation | A numeric table of size  that contains the computed out-of-bag error
when the computeOutOfBagErrorPerObservation option is enabled. The value
 in the table indicates that no OOB value was computed because this
observation was not in OOB set for any of the trees in the model (never
left out during the bootstrap).By default, this result is an object of the HomogenNumericTable class,
but you can define the result as an object of any class derived from NumericTable . |
updatedEngine | Engine instance with state updated after computations. |
Performance Considerations
To get the best performance of the decision forest variable
importance computation, use the Mean Decrease Impurity (MDI) rather
than the Mean Decrease Accuracy (MDA) method.
Product and Performance Information |
|---|
Performance varies by use, configuration and other factors.
Learn more at www.Intel.com/PerformanceIndex. Notice revision #20201201 |