DecisionTreeClassifier constructor
Parameters:
trainData
A DataFrame
with observations that will be used to build a
decision tree. Must contain targetName
column.
targetName
A name of a column in trainData
that contains class labels
minError
A value within the range 0..1 (both inclusive). The value is a
minimal error on a single decision tree node and is used as a stop
criterion to avoid further decision tree node splitting: if the node is
good enough, there is no need to split it and thus it will become a leaf.
minSamplesCount
A minimal number of samples (observations) on the
decision's tree node. The value is used as a stop criteria to avoid
further decision tree node splitting: if the node contains less than or
equal to minSamplesCount
observations, the node turns into the leaf.
maxDepth
A maximum number of decision tree levels.
assessorType
Defines an assessment type that will be applied to the
data in order to decide how to split the subset while building the tree.
Default value is TreeAssessorType.gini
Possible values of assessorType
:
TreeAssessorType.gini The algorithm makes a decision on how to split a subset of data based on the Gini index
TreeAssessorType.majority The algorithm makes a decision on how to split a subset of data based on a major class.
Implementation
factory DecisionTreeClassifier(
DataFrame trainData,
String targetName, {
num minError = 0.5,
int minSamplesCount = 1,
int maxDepth = 10,
DType dtype = dTypeDefaultValue,
TreeAssessorType assessorType = TreeAssessorType.gini,
}) =>
initDecisionTreeModule().get<DecisionTreeClassifierFactory>().create(
trainData,
targetName,
dtype,
minError,
minSamplesCount,
maxDepth,
assessorType,
);