Implementation:Online ml River Tree Splitter Gaussian

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

Overview

Gaussian-based attribute observer for classification that approximates class distributions using Gaussian estimators for probability density calculation.

Description

GaussianSplitter approximates the distribution of each class for a numerical feature using Gaussian (normal) distributions. This enables efficient probability density function calculation for Naive Bayes predictions. The splitter tracks minimum and maximum values per class and suggests split candidates by partitioning the feature range into equal-sized bins. Split evaluation uses the CDF of Gaussian distributions to estimate class distributions.

Usage

Use GaussianSplitter for classification tasks when features approximately follow normal distributions within each class and when Naive Bayes leaf models are used.

Code Reference

Source Location

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

Signature

class GaussianSplitter(Splitter):
    def __init__(self, n_splits: int = 10):
        ...

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

    def cond_proba(self, att_val, target_val):
        ...

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

Import

from river.tree.splitter import GaussianSplitter

I/O Contract

Input	Type	Description
att_val	float	Numerical feature value
target_val	int/str	Class label
w	float	Sample weight
n_splits	int	Number of split candidates to evaluate (default 10)

Output	Type	Description
cond_proba	float	class) from Gaussian PDF
split_suggestion	BranchFactory	Best split with estimated post-split distributions

Usage Examples

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

# Create splitter with 15 split candidates
splitter = GaussianSplitter(n_splits=15)

# Update with observations
splitter.update(5.5, 'cat', 1.0)
splitter.update(6.2, 'dog', 1.0)
splitter.update(5.8, 'cat', 1.0)
splitter.update(7.1, 'dog', 1.0)

# Get conditional probability for Naive Bayes
prob = splitter.cond_proba(att_val=5.5, target_val='cat')
print(f"P(5.5 | cat) = {prob}")

# Get best split
criterion = GiniSplitCriterion()
pre_split = {'cat': 100, 'dog': 80}

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

Related Pages

• Environment:Online_ml_River_Python_Runtime_Environment