biz.sc.mm.dm.modelstats.tree

Interface Split

All Superinterfaces:

Verifiable

public interface Split
extends Verifiable

Contains all information for split on a single attribute: possible split values, and train and validation data statistics for each value.

Field Summary

Fields

Modifier and Type	Field and Description
static int	NO_SUBSET_FOR_NULLS If there is not distinguished subset for null cases.
static int	SUBSET_FOR_NULLS_NOT_SET If the subset's index has not been set yet.

Method Summary

Methods

Modifier and Type	Method and Description
SplitValue	getActualSplit() The returned split is the actual (current) split which currently works while using the tree for viewing, applying, testing etc.
java.lang.String	getAttributeName() Returns split attribute name.
SplitValue	getBuiltSplit() The returned split is the original split in model created by build task.
ComparisonOperator	getComparisonOperator(int childIndx) Gets the split operator for i'th child.
java.lang.String	getLabel(int i) Gets the split label for i'th child.
SplitValue	getOptimalSplit() The returned split is the split created by algorithm, that means the split is optimal using split evaluation criterion..
java.lang.String	getRule(int i) Gets the predicate rule for i'th child.
double	getSplitEvaluation() Returns evaluation value for this split, if computed in the tree building phase.
int	getSubsetForNulls() Which subset gets the null cases.
SplitStats	getTrain() Returns split statistics for train data.
SplitStats	getValidation() Returns split statistics for validation data.
boolean	isNumerical() Numerical split type flag.
void	setActualSplit(SplitValue val) The returned split is the actual split which currently works while using the tree for viewing, applying, testing etc.
void	setPredicates(Predicate[] predicates) Sets children predicates
void	setSubsetForNulls(int subsetForNulls) Sets which subset is going to be used to process null cases.

Methods inherited from interface biz.sc.mm.dm.base.Verifiable

verify

Field Detail

NO_SUBSET_FOR_NULLS

static final int NO_SUBSET_FOR_NULLS

If there is not distinguished subset for null cases.

See Also:

Constant Field Values

SUBSET_FOR_NULLS_NOT_SET

static final int SUBSET_FOR_NULLS_NOT_SET

If the subset's index has not been set yet.

See Also:

Constant Field Values

Method Detail

getActualSplit

SplitValue getActualSplit()

The returned split is the actual (current) split which currently works while using the tree for viewing, applying, testing etc. After build task, this is equal to getBuiltSplit() and can be manually changed.

Returns:

threshold value for this binary split

getAttributeName

java.lang.String getAttributeName()

Returns split attribute name.

Returns:

getBuiltSplit

SplitValue getBuiltSplit()

The returned split is the original split in model created by build task. That means the split is either a split modified manually in previous model, or optimal using split evaluation criterion

Returns:

threshold value for this binary split

getComparisonOperator

ComparisonOperator getComparisonOperator(int childIndx)

Gets the split operator for i'th child.

Parameters:

childIndx -

Returns:

getLabel

java.lang.String getLabel(int i)

Gets the split label for i'th child.

getOptimalSplit

SplitValue getOptimalSplit()

The returned split is the split created by algorithm, that means the split is optimal using split evaluation criterion..

Returns:

threshold value for this binary split

getRule

java.lang.String getRule(int i)

Gets the predicate rule for i'th child.

Parameters:

i -

Returns:

getSplitEvaluation

double getSplitEvaluation()

Returns evaluation value for this split, if computed in the tree building phase. The property is optional. The value refers to the split from the parent node.

Returns:

evaluation value for this predicate, or NaN if not apriopriate.

getTrain

SplitStats getTrain()

Returns split statistics for train data.

Returns:

getValidation

SplitStats getValidation()

Returns split statistics for validation data.

Returns:

isNumerical

boolean isNumerical()

Numerical split type flag. If the split is numerical, than set to true.

Returns:

numerical split type flag.

setActualSplit

void setActualSplit(SplitValue val)

The returned split is the actual split which currently works while using the tree for viewing, applying, testing etc. After build task, it is equal to getBuiltSplit() and can be manually changed.

Parameters:

val - threshold value for this binary split

setPredicates

void setPredicates(Predicate[] predicates)

Sets children predicates

Parameters:

predicates -

getSubsetForNulls

int getSubsetForNulls()

Which subset gets the null cases.

 -1 if there is not distinguished subset for null cases.  
 -2 if the subset's index has not been set yet.

setSubsetForNulls

void setSubsetForNulls(int subsetForNulls)

Sets which subset is going to be used to process null cases.

Parameters:

subsetForNulls - The subset's index.