tests/test_modeling.py

from __future__ import annotations

import warnings

import pandas as pd

from sepsis_mcp.modeling import ProbabilityEstimator


def test_probability_estimator_supports_logistic_regression() -> None:
    features = pd.DataFrame(
        {
            "feature_a": [0.0, 0.2, 0.8, 1.0, 0.1, 0.9],
            "feature_b": [0.0, 0.1, 0.9, 1.0, 0.2, 0.8],
        }
    )
    labels = pd.Series([0, 0, 1, 1, 0, 1], dtype=int)

    estimator = ProbabilityEstimator(random_state=0, model_type="logistic_regression")
    estimator.fit(features, labels)
    probabilities = estimator.predict_positive_proba(features)

    assert probabilities.between(0.0, 1.0).all()
    assert len(probabilities) == len(features)


def test_probability_estimator_logistic_regression_handles_nan_features() -> None:
    features = pd.DataFrame(
        {
            "feature_a": [0.0, float("nan"), 0.8, 1.0, 0.1, 0.9],
            "feature_b": [0.0, 0.1, float("nan"), 1.0, 0.2, 0.8],
            "feature_c": [float("nan")] * 6,
        }
    )
    labels = pd.Series([0, 0, 1, 1, 0, 1], dtype=int)

    with warnings.catch_warnings(record=True) as caught:
        warnings.simplefilter("always")
        estimator = ProbabilityEstimator(random_state=0, model_type="logistic_regression")
        estimator.fit(features, labels)
    probabilities = estimator.predict_positive_proba(features)

    assert caught == []
    assert probabilities.between(0.0, 1.0).all()


def test_probability_estimator_supports_mlp() -> None:
    features = pd.DataFrame(
        {
            "feature_a": [0.0, 0.2, 0.8, 1.0, 0.1, 0.9],
            "feature_b": [0.0, 0.1, 0.9, 1.0, 0.2, 0.8],
            "feature_c": [float("nan"), 0.3, 0.7, float("nan"), 0.4, 0.6],
        }
    )
    labels = pd.Series([0, 0, 1, 1, 0, 1], dtype=int)

    estimator = ProbabilityEstimator(random_state=0, model_type="mlp")
    estimator.fit(features, labels)
    probabilities = estimator.predict_positive_proba(features)

    assert probabilities.between(0.0, 1.0).all()
    assert len(probabilities) == len(features)