Implementation:Online ml River Tree Nodes EFDTC

Knowledge Sources	Online_ml_River
Domains	Online_Learning, Decision_Trees
Last Updated	2026-02-08 16:00 GMT

Overview

EFDT (Extremely Fast Decision Tree) node implementations for classification tasks that support both learning and splitting operations simultaneously.

Description

This module implements the specialized nodes for EFDT classifiers in River. EFDT is unique among Hoeffding Trees because split nodes continue learning from incoming samples and can reevaluate split decisions over time. The module provides both leaf nodes (EFDTLeafMajorityClass, EFDTLeafNaiveBayes, EFDTLeafNaiveBayesAdaptive) and branch nodes (EFDTNominalBinaryBranch, EFDTNominalMultiwayBranch, EFDTNumericBinaryBranch, EFDTNumericMultiwayBranch) that inherit from base classes BaseEFDTLeaf and BaseEFDTBranch.

Usage

Use EFDT nodes when building extremely fast decision tree classifiers that need to adapt quickly to concept drift by revisiting split decisions. These nodes are ideal for high-speed data streams where rapid model updates are critical.

Code Reference

Source Location

• Repository: Online_ml_River
• File: river/tree/nodes/efdtc_nodes.py

Signature

class BaseEFDTLeaf(HTLeaf):
    def __init__(self, stats, depth, splitter, **kwargs):
        ...
    def best_split_suggestions(self, criterion, tree):
        ...

class BaseEFDTBranch(DTBranch):
    def __init__(self, stats, *children, splitter, splitters, **attributes):
        ...
    def learn_one(self, x, y, *, w=1.0, tree=None):
        ...
    def best_split_suggestions(self, criterion, tree):
        ...

class EFDTLeafMajorityClass(BaseEFDTLeaf, LeafMajorityClass):
    ...

class EFDTLeafNaiveBayes(BaseEFDTLeaf, LeafNaiveBayes):
    ...

class EFDTLeafNaiveBayesAdaptive(BaseEFDTLeaf, LeafNaiveBayesAdaptive):
    ...

Import

from river.tree.nodes.efdtc_nodes import EFDTLeafMajorityClass
from river.tree.nodes.efdtc_nodes import EFDTLeafNaiveBayes
from river.tree.nodes.efdtc_nodes import BaseEFDTBranch

I/O Contract

Input	Type	Description
stats	dict	Class observations dictionary
depth	int	Node depth in the tree
splitter	Splitter	Numeric attribute observer for monitoring statistics

Output	Type	Description
prediction	dict	Normalized class probability distribution
split_suggestions	list[BranchFactory]	Candidate splits for the node

Usage Examples

from river.tree.nodes.efdtc_nodes import EFDTLeafMajorityClass
from river.tree.splitter import GaussianSplitter

# Create an EFDT leaf node
leaf = EFDTLeafMajorityClass(
    stats={0: 10, 1: 15},
    depth=2,
    splitter=GaussianSplitter()
)

# Update the leaf with new samples
x = {'feature1': 0.5, 'feature2': 1.2}
leaf.learn_one(x, y=1, w=1.0, tree=None)

# Get predictions
prediction = leaf.prediction(x, tree=None)

# EFDT branches can also learn
from river.tree.nodes.efdtc_nodes import EFDTNumericBinaryBranch
# Branches continue learning as data passes through

Related Pages

• Environment:Online_ml_River_Python_Runtime_Environment