Implementation:Online ml River Stats Mode

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

Overview

Mode tracks the most frequently occurring value in a data stream.

Description

This statistic identifies the most common value (mode) in streaming data by maintaining a counter of observed values. It supports both exact mode computation and approximate mode by limiting the number of unique values tracked (controlled by parameter k). The implementation includes Mode for overall mode and RollingMode for mode within a sliding window, both using efficient counter-based tracking.

Usage

Use Mode when you need to identify the most common value in categorical or discrete streaming data. Common applications include finding the most frequent category in classification problems, identifying typical values in categorical features, detecting prevalent patterns, and summarizing categorical distributions where the median or mean are not applicable.

Code Reference

Source Location

• Repository: Online_ml_River
• File: river/stats/mode.py

Signature

class Mode(stats.base.Univariate):
    def __init__(self, k=25):
        self.k = k
        self.counts = collections.defaultdict(int)

class RollingMode(stats.base.RollingUnivariate):
    def __init__(self, window_size: int):
        self.window: collections.deque[numbers.Number] = collections.deque(maxlen=window_size)
        self.counts: collections.defaultdict[typing.Any, int] = collections.defaultdict(int)

Import

from river import stats

I/O Contract

Inputs

Name	Type	Required	Description
x	Any (hashable)	Yes	Value to update the statistic with
k	int	Yes (init)	Number of unique values to track (-1 for exact mode, default: 25)
window_size	int	Yes (for Rolling)	Size of the rolling window

Outputs

Name	Type	Description
get()	Any	Most frequently occurring value (or None if no values seen)

Usage Examples

from river import stats

# Mode with limited unique values
X = ['sunny', 'cloudy', 'cloudy', 'rainy', 'rainy', 'rainy']
mode = stats.Mode(k=2)

for x in X:
    mode.update(x)
    print(f"Value: {x}, Mode: {mode.get()}")

# Output:
# Value: sunny, Mode: sunny
# Value: cloudy, Mode: sunny
# Value: cloudy, Mode: cloudy
# Value: rainy, Mode: cloudy  (rainy not tracked due to k=2)
# Value: rainy, Mode: cloudy
# Value: rainy, Mode: cloudy

# Exact mode computation
mode_exact = stats.Mode(k=-1)
for x in X:
    mode_exact.update(x)
    print(f"Value: {x}, Exact Mode: {mode_exact.get()}")

# Output:
# Value: sunny, Mode: sunny
# Value: cloudy, Mode: sunny
# Value: cloudy, Mode: cloudy
# Value: rainy, Mode: cloudy
# Value: rainy, Mode: cloudy
# Value: rainy, Mode: rainy

# Rolling mode
X = ['sunny', 'sunny', 'sunny', 'rainy', 'rainy', 'rainy', 'rainy']
rolling_mode = stats.RollingMode(window_size=2)

for x in X:
    rolling_mode.update(x)
    print(f"Value: {x}, Rolling Mode: {rolling_mode.get()}")

# Output:
# Value: sunny, Rolling Mode: sunny
# Value: sunny, Rolling Mode: sunny
# Value: sunny, Rolling Mode: sunny
# Value: rainy, Rolling Mode: sunny
# Value: rainy, Rolling Mode: rainy
# Value: rainy, Rolling Mode: rainy
# Value: rainy, Rolling Mode: rainy

# Numeric mode
numeric_mode = stats.Mode()
for x in [1, 2, 2, 3, 3, 3]:
    numeric_mode.update(x)
print(f"Most frequent number: {numeric_mode.get()}")
# Output: 3

Related Pages

• Environment:Online_ml_River_Python_Runtime_Environment