swarmauri_measurement_mutualinformation 0.11.0.dev1

Mutual-information feature screening measurement for Swarmauri classification datasets using scikit-learn.

Swarmauri Measurement Mutual Information

swarmauri_measurement_mutualinformation is the Swarmauri feature-signal measurement for supervised classification datasets. It wraps sklearn.feature_selection.mutual_info_classif and returns the average mutual information across all non-target columns in a Pandas DataFrame.

Why Use Swarmauri Measurement Mutual Information

• Estimate how strongly each feature depends on a discrete target before model training.
• Reduce low-signal columns earlier in a Swarmauri data or evaluation pipeline.
• Reuse a standard MeasurementBase component instead of hand-wiring feature scoring logic.
• Pair feature screening with downstream metrics, vectorization, or model selection flows.

FAQ

What input does this measurement expect?
A Pandas DataFrame plus the name of the target column.

What does it return?
A single float: the mean of the per-feature mutual-information scores produced by mutual_info_classif.

Does it return one score per feature?
No. This component averages the feature scores. If you need per-feature values, call mutual_info_classif directly.

What units are the underlying scores in?
Scikit-learn documents mutual_info_classif outputs in natural-log units.

Features

• Supervised mutual-information scoring for discrete targets.
• Automatic exclusion of the target column from the feature matrix.
• Returns one aggregate score that is easy to log, compare, or threshold.
• Uses the Swarmauri measurement interface for pipeline compatibility.
• Supports Python 3.10, 3.11, 3.12, 3.13, and 3.14.

Installation

uv add swarmauri_measurement_mutualinformation

pip install swarmauri_measurement_mutualinformation

Usage

import pandas as pd
from swarmauri_measurement_mutualinformation import MutualInformationMeasurement

frame = pd.DataFrame(
    {
        "feature_a": [0, 1, 1, 0, 1, 0],
        "feature_b": [5.1, 5.0, 4.9, 5.2, 5.1, 5.0],
        "target": [0, 1, 1, 0, 1, 0],
    }
)

measurement = MutualInformationMeasurement()
score = measurement.calculate(frame, target_column="target")
print(score)

Examples

Screen a small classification dataset

import pandas as pd
from swarmauri_measurement_mutualinformation import MutualInformationMeasurement

data = pd.DataFrame(
    {
        "clicked_email": [1, 0, 1, 1, 0, 0],
        "days_active": [4, 2, 5, 6, 1, 2],
        "plan_tier": [2, 1, 2, 3, 1, 1],
        "converted": [1, 0, 1, 1, 0, 0],
    }
)

measurement = MutualInformationMeasurement()
print(measurement.calculate(data, target_column="converted"))

Inspect per-feature scores directly

from sklearn.feature_selection import mutual_info_classif

X = frame.drop(columns=["target"])
y = frame["target"]

scores = mutual_info_classif(X, y)
for column, score in zip(X.columns, scores):
    print(column, score)

Related Packages

• swarmauri_measurement_tokencountestimator
• swarmauri_metric_hamming
• swarmauri_measurement_mutualinformation

Swarmauri Foundations

• swarmauri
• swarmauri_core
• swarmauri_base

More Documentation

• scikit-learn mutual_info_classif docs
• Pandas documentation

Best Practices

• Encode categorical columns numerically before calling this measurement.
• Remove or impute missing values before scoring.
• Use a stable preprocessing pipeline when comparing MI across experiments.
• Inspect the raw per-feature scores if feature-level ranking matters more than a single aggregate summary.

License

This project is licensed under the Apache-2.0 License.