kumulant

InformationGain

@Serializable
@SerialName(value = "InformationGain")
data object InformationGain : ClassificationSplitMetric(source)

Information-gain split criterion: parent entropy minus weighted children entropy.

Functions

Link copied to clipboard
open override fun score(total: ClassCountsResult, pos: ClassCountsResult, neg: ClassCountsResult): Double

Score a candidate split given the leaf's total / pos / neg class counts.

score

Score a candidate split given the leaf's total / pos / neg class counts.