Spaces:
Sleeping
Sleeping
File size: 3,921 Bytes
f7d11f7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 |
import joblib
import numpy as np
import pandas as pd
from predicting_outcomes_in_heart_failure.config import (
MODELS_DIR,
PREPROCESSED_CSV,
TARGET_COL,
)
import pytest
@pytest.fixture
def sample_features():
df = pd.read_csv(PREPROCESSED_CSV).iloc[100:].copy()
return df.iloc[[1]].drop(columns=[TARGET_COL])
@pytest.fixture
def trained_models():
"""
Load all saved models.
"""
models = {}
models_path = MODELS_DIR / "all"
for model_file in models_path.iterdir():
if model_file.suffix == ".joblib":
model_name = model_file.stem
models[model_name] = joblib.load(model_file)
return models
class TestModelMinimumFunctionality:
def test_predict_returns_binary_array(self, trained_models, sample_features):
"""
Models should return binary predictions (0 or 1) for heart disease.
"""
models = trained_models
for model_name, model in models.items():
predictions = model.predict(sample_features)
assert predictions.shape == (1,), f"{model_name}: wrong shape"
assert predictions[0] in [0, 1], f"{model_name}: prediction not binary"
def test_predict_proba_returns_valid_probabilities(self, trained_models, sample_features):
"""
Models with predict_proba should return valid probability distributions.
"""
models = trained_models
for model_name, model in models.items():
if hasattr(model, "predict_proba"):
proba = model.predict_proba(sample_features)
assert proba.shape == (1, 2), f"{model_name}: wrong probability shape"
assert np.allclose(proba.sum(axis=1), 1.0), (
f"{model_name}: probabilities don't sum to 1"
)
assert np.all(proba >= 0) and np.all(proba <= 1), (
f"{model_name}: invalid probability values"
)
def test_batch_prediction_consistency(self, trained_models, sample_features):
"""
Predictions on batches should be consistent with individual predictions.
"""
models = trained_models
batch = pd.concat([sample_features] * 5, ignore_index=True)
for model_name, model in models.items():
batch_pred = model.predict(batch)
single_pred = model.predict(sample_features)
assert len(batch_pred) == 5, f"{model_name}: wrong batch size"
assert np.all(batch_pred == single_pred[0]), (
f"{model_name}: inconsistent batch predictions"
)
def test_correct_number_of_features(self, trained_models, sample_features):
"""
Models should expect exactly 18 features after preprocessing
"""
models = trained_models
expected_n_features = 18
assert len(sample_features.columns) == expected_n_features, (
f"Sample has {len(sample_features.columns)} columns, expected {expected_n_features}"
)
for _, model in models.items():
_ = model.predict(sample_features)
with pytest.raises((ValueError, IndexError)):
wrong_features = sample_features.iloc[:, :10]
model.predict(wrong_features)
def test_invalid_one_hot_encoding_detection(self, trained_models, sample_features):
"""
Test behavior when one-hot encoding is invalid
"""
models = trained_models
invalid_none = sample_features.copy()
invalid_none["ChestPainType_ASY"] = 0
invalid_none["ChestPainType_ATA"] = 0
invalid_none["ChestPainType_NAP"] = 0
invalid_none["ChestPainType_TA"] = 0
for model_name, model in models.items():
pred = model.predict(invalid_none)
assert pred[0] in [0, 1], f"{model_name}: should still return valid binary prediction"
|