Implementation:Online ml River Metrics GeometricMean
| Knowledge Sources | |
|---|---|
| Domains | Online_Learning, Evaluation_Metrics |
| Last Updated | 2026-02-08 16:00 GMT |
Overview
Geometric mean of class-wise sensitivity (recall) for imbalanced classification evaluation.
Description
GeometricMean computes the nth root of the product of class-wise sensitivities (recalls), where n is the number of classes. The formula is gm = (s1 × s2 × ... × sn)^(1/n), where si is the sensitivity of class i. This metric is particularly effective for imbalanced datasets as it's independent of class distribution and poor performance on any single class significantly lowers the overall score.
Usage
Use GeometricMean when evaluating classifiers on imbalanced datasets where all classes are equally important. Unlike arithmetic averages, the geometric mean severely penalizes poor performance on minority classes, making it ideal for scenarios where you cannot afford to ignore any class. A classifier must perform well across all classes to achieve a high geometric mean score.
Code Reference
Source Location
- Repository: Online_ml_River
- File: river/metrics/geometric_mean.py
Signature
class GeometricMean(metrics.base.MultiClassMetric):
def __init__(self, cm=None):
pass
Import
from river import metrics
I/O Contract
| Method | Parameters | Returns | Description |
|---|---|---|---|
| update | y_true, y_pred, [w] | None | Updates metric with true and predicted labels |
| get | - | float | Returns geometric mean of class sensitivities (0.0 to 1.0) |
Usage Examples
from river import metrics
y_true = ['cat', 'ant', 'cat', 'cat', 'ant', 'bird', 'bird']
y_pred = ['ant', 'ant', 'cat', 'cat', 'ant', 'cat', 'bird']
metric = metrics.GeometricMean()
for yt, yp in zip(y_true, y_pred):
metric.update(yt, yp)
print(metric)
# GeometricMean: 69.34%
# The metric computes:
# Sensitivity(cat) = 2/3 = 66.67%
# Sensitivity(ant) = 2/2 = 100%
# Sensitivity(bird) = 1/2 = 50%
# GeometricMean = (0.6667 × 1.0 × 0.5)^(1/3) = 0.6934
# Note how the low sensitivity on 'bird' pulls down
# the overall score more than an arithmetic mean would