Implementation:Online ml River Tree Splitter Exhaustive

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

Overview

Exhaustive attribute observer for classification that stores all observations in a binary search tree to evaluate all possible split points.

Description

ExhaustiveSplitter implements the original attribute observer from Domingos and Hulten's VFDT algorithm. It uses a binary search tree to store all feature values and class counts between split attempts, enabling exhaustive evaluation of split candidates. Each node stores left and right class distributions. This splitter cannot perform probability density estimation and does not work well with Naive Bayes leaf models.

Usage

Use ExhaustiveSplitter for classification tasks when you want to evaluate all possible split points. It provides thorough split evaluation but has higher memory requirements than approximate methods.

Code Reference

Source Location

• Repository: Online_ml_River
• File: river/tree/splitter/exhaustive_splitter.py

Signature

class ExhaustiveSplitter(Splitter):
    def __init__(self):
        ...

    def update(self, att_val, target_val, w):
        ...

    def cond_proba(self, att_val, target_val):
        return 0.0

    def best_evaluated_split_suggestion(self, criterion, pre_split_dist, att_idx, binary_only):
        ...


class ExhaustiveNode:
    def __init__(self, att_val, target_val, w):
        ...

    def insert_value(self, val, label, w):
        ...

Import

from river.tree.splitter import ExhaustiveSplitter

I/O Contract

Input	Type	Description
att_val	float	Numerical feature value
target_val	int/str	Class label
w	float	Sample weight
criterion	SplitCriterion	Split evaluation criterion

Output	Type	Description
split_suggestion	BranchFactory	Best split with merit and post-split distributions
cond_proba	float	Always 0.0 (no density estimation)

Usage Examples

from river.tree.splitter import ExhaustiveSplitter
from river.tree.split_criterion import GiniSplitCriterion

splitter = ExhaustiveSplitter()

# Update with observations
splitter.update(5.5, 'A', 1.0)
splitter.update(6.2, 'B', 1.0)
splitter.update(4.8, 'A', 1.0)
splitter.update(7.1, 'B', 1.0)

# Get best split
criterion = GiniSplitCriterion()
pre_split = {'A': 10, 'B': 8}

suggestion = splitter.best_evaluated_split_suggestion(
    criterion=criterion,
    pre_split_dist=pre_split,
    att_idx='feature1',
    binary_only=True
)

print(f"Best threshold: {suggestion.split_info}")
print(f"Merit: {suggestion.merit}")
print(f"Left distribution: {suggestion.children_stats[0]}")
print(f"Right distribution: {suggestion.children_stats[1]}")

Related Pages

• Environment:Online_ml_River_Python_Runtime_Environment