Implementation:Online ml River Tree Splitter NominalClassif

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

Overview

Splitter for nominal (categorical) features in classification tasks using dictionary-based class count tracking.

Description

NominalSplitterClassif monitors categorical features by maintaining class counts for each feature value using dictionary structures. It supports both binary splits (testing equality to a specific value) and multiway splits (creating one branch per category). The splitter tracks all observed feature values and missing value weights, enabling conditional probability estimation for Naive Bayes.

Usage

Use NominalSplitterClassif when monitoring categorical features in classification trees. It automatically handles both binary and multiway split evaluation.

Code Reference

Source Location

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

Signature

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

    @property
    def is_numeric(self):
        return False

    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 NominalSplitterClassif

I/O Contract

Input	Type	Description
att_val	any	Categorical feature value
target_val	int/str	Class label
w	float	Sample weight
binary_only	bool	If True, only binary splits; else include multiway

Output	Type	Description
cond_proba	float	class)
split_suggestion	BranchFactory	Best split (binary or multiway)

Usage Examples

from river.tree.splitter.nominal_splitter_classif import NominalSplitterClassif
from river.tree.split_criterion import GiniSplitCriterion

splitter = NominalSplitterClassif()

# Update with categorical observations
splitter.update('red', 'positive', 1.0)
splitter.update('blue', 'negative', 1.0)
splitter.update('red', 'positive', 1.0)
splitter.update('green', 'negative', 1.0)

# Get conditional probability
prob = splitter.cond_proba('red', 'positive')
print(f"P(red | positive) = {prob}")

# Get best binary split
criterion = GiniSplitCriterion()
pre_split = {'positive': 100, 'negative': 80}

binary_split = splitter.best_evaluated_split_suggestion(
    criterion=criterion,
    pre_split_dist=pre_split,
    att_idx='color',
    binary_only=True
)

# Get best multiway split
multiway_split = splitter.best_evaluated_split_suggestion(
    criterion=criterion,
    pre_split_dist=pre_split,
    att_idx='color',
    binary_only=False
)

Related Pages

• Environment:Online_ml_River_Python_Runtime_Environment