Skip to content

glassbox.metrics

Evaluation metrics for classification and regression.

accuracy_score 

accuracy_score(y_true, y_pred)

Compute the classification accuracy.

Parameters:

Name Type Description Default
y_true ndarray

Ground truth class labels of shape (n_samples,).

 required
y_pred ndarray

Predicted class labels of shape (n_samples,).

 required

Returns:

Type Description
float

Classification accuracy in the range [0.0, 1.0].
Source code in glassbox/metrics/classification.py 
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
def accuracy_score(y_true: np.ndarray, y_pred: np.ndarray) -> float:
    """
    Compute the classification accuracy.

    Parameters
    ----------
    y_true : np.ndarray
        Ground truth class labels of shape (n_samples,).
    y_pred : np.ndarray
        Predicted class labels of shape (n_samples,).

    Returns
    -------
    float
        Classification accuracy in the range [0.0, 1.0].
    """
    y_true_arr, y_pred_arr = _validate_classification_inputs(y_true=y_true, y_pred=y_pred)
    return float(np.mean(y_true_arr == y_pred_arr))

confusion_matrix 

confusion_matrix(y_true, y_pred)

Compute the confusion matrix for classification results.

Parameters:

Name Type Description Default
y_true ndarray

Ground truth class labels of shape (n_samples,).

 required
y_pred ndarray

Predicted class labels of shape (n_samples,).

 required

Returns:

Type Description
ndarray

Confusion matrix of shape (n_classes, n_classes).
Notes

Rows correspond to true labels and columns correspond to predicted labels. Class order follows sorted unique labels from the union of y_true and y_pred.

Source code in glassbox/metrics/classification.py 
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
def confusion_matrix(y_true: np.ndarray, y_pred: np.ndarray) -> np.ndarray:
    """
    Compute the confusion matrix for classification results.

    Parameters
    ----------
    y_true : np.ndarray
        Ground truth class labels of shape (n_samples,).
    y_pred : np.ndarray
        Predicted class labels of shape (n_samples,).

    Returns
    -------
    np.ndarray
        Confusion matrix of shape (n_classes, n_classes).

    Notes
    -----
    Rows correspond to true labels and columns correspond to predicted labels.
    Class order follows sorted unique labels from the union of `y_true` and
    `y_pred`.
    """
    y_true_arr, y_pred_arr = _validate_classification_inputs(y_true=y_true, y_pred=y_pred)

    labels = np.unique(np.concatenate((y_true_arr, y_pred_arr)))
    label_to_index = {label: idx for idx, label in enumerate(labels)}

    matrix = np.zeros((labels.shape[0], labels.shape[0]), dtype=int)
    for true_label, pred_label in zip(y_true_arr, y_pred_arr):
        true_idx = label_to_index[true_label]
        pred_idx = label_to_index[pred_label]
        matrix[true_idx, pred_idx] += 1

    return matrix

f1_score 

f1_score(y_true, y_pred)

Compute the F1 score for classification.

Parameters:

Name Type Description Default
y_true ndarray

Ground truth class labels of shape (n_samples,).

 required
y_pred ndarray

Predicted class labels of shape (n_samples,).

 required

Returns:

Type Description
float

F1 score in the range [0.0, 1.0].
Notes

Uses macro averaging over all classes with per-class zero-division handling.

Source code in glassbox/metrics/classification.py 
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
def f1_score(y_true: np.ndarray, y_pred: np.ndarray) -> float:
    """
    Compute the F1 score for classification.

    Parameters
    ----------
    y_true : np.ndarray
        Ground truth class labels of shape (n_samples,).
    y_pred : np.ndarray
        Predicted class labels of shape (n_samples,).

    Returns
    -------
    float
        F1 score in the range [0.0, 1.0].

    Notes
    -----
    Uses macro averaging over all classes with per-class zero-division handling.
    """
    cm = confusion_matrix(y_true=y_true, y_pred=y_pred)

    true_positives = np.diag(cm).astype(float)
    predicted_positives = np.sum(cm, axis=0).astype(float)
    actual_positives = np.sum(cm, axis=1).astype(float)

    precision_per_class = _safe_divide(true_positives, predicted_positives)
    recall_per_class = _safe_divide(true_positives, actual_positives)

    f1_numerator = 2.0 * precision_per_class * recall_per_class
    f1_denominator = precision_per_class + recall_per_class
    f1_per_class = _safe_divide(f1_numerator, f1_denominator)
    return float(np.mean(f1_per_class))

precision_score 

precision_score(y_true, y_pred)

Compute the classification precision score.

Parameters:

Name Type Description Default
y_true ndarray

Ground truth class labels of shape (n_samples,).

 required
y_pred ndarray

Predicted class labels of shape (n_samples,).

 required

Returns:

Type Description
float

Precision score in the range [0.0, 1.0].
Notes

Uses macro averaging over all classes. Classes with zero predicted samples contribute a precision of 0.0.

Source code in glassbox/metrics/classification.py 
 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
def precision_score(y_true: np.ndarray, y_pred: np.ndarray) -> float:
    """
    Compute the classification precision score.

    Parameters
    ----------
    y_true : np.ndarray
        Ground truth class labels of shape (n_samples,).
    y_pred : np.ndarray
        Predicted class labels of shape (n_samples,).

    Returns
    -------
    float
        Precision score in the range [0.0, 1.0].

    Notes
    -----
    Uses macro averaging over all classes. Classes with zero predicted samples
    contribute a precision of 0.0.
    """
    cm = confusion_matrix(y_true=y_true, y_pred=y_pred)
    true_positives = np.diag(cm).astype(float)
    predicted_positives = np.sum(cm, axis=0).astype(float)
    precision_per_class = _safe_divide(true_positives, predicted_positives)
    return float(np.mean(precision_per_class))

recall_score 

recall_score(y_true, y_pred)

Compute the classification recall score.

Parameters:

Name Type Description Default
y_true ndarray

Ground truth class labels of shape (n_samples,).

 required
y_pred ndarray

Predicted class labels of shape (n_samples,).

 required

Returns:

Type Description
float

Recall score in the range [0.0, 1.0].
Notes

Uses macro averaging over all classes. Classes with zero true samples contribute a recall of 0.0.

Source code in glassbox/metrics/classification.py 
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
def recall_score(y_true: np.ndarray, y_pred: np.ndarray) -> float:
    """
    Compute the classification recall score.

    Parameters
    ----------
    y_true : np.ndarray
        Ground truth class labels of shape (n_samples,).
    y_pred : np.ndarray
        Predicted class labels of shape (n_samples,).

    Returns
    -------
    float
        Recall score in the range [0.0, 1.0].

    Notes
    -----
    Uses macro averaging over all classes. Classes with zero true samples
    contribute a recall of 0.0.
    """
    cm = confusion_matrix(y_true=y_true, y_pred=y_pred)
    true_positives = np.diag(cm).astype(float)
    actual_positives = np.sum(cm, axis=1).astype(float)
    recall_per_class = _safe_divide(true_positives, actual_positives)
    return float(np.mean(recall_per_class))

mean_absolute_error 

mean_absolute_error(y_true, y_pred)

Compute the mean absolute error (MAE).

Parameters:

Name Type Description Default
y_true ndarray

Ground truth target values of shape (n_samples,).

 required
y_pred ndarray

Predicted target values of shape (n_samples,).

 required

Returns:

Type Description
float

Mean absolute error.
Source code in glassbox/metrics/regression.py 
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
def mean_absolute_error(y_true: np.ndarray, y_pred: np.ndarray) -> float:
    """
    Compute the mean absolute error (MAE).

    Parameters
    ----------
    y_true : np.ndarray
        Ground truth target values of shape (n_samples,).
    y_pred : np.ndarray
        Predicted target values of shape (n_samples,).

    Returns
    -------
    float
        Mean absolute error.
    """
    y_true_arr, y_pred_arr = _validate_regression_inputs(y_true=y_true, y_pred=y_pred)
    return float(np.mean(np.abs(y_true_arr - y_pred_arr)))

mean_squared_error 

mean_squared_error(y_true, y_pred)

Compute the mean squared error (MSE).

Parameters:

Name Type Description Default
y_true ndarray

Ground truth target values of shape (n_samples,).

 required
y_pred ndarray

Predicted target values of shape (n_samples,).

 required

Returns:

Type Description
float

Mean squared error.
Source code in glassbox/metrics/regression.py 
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
def mean_squared_error(y_true: np.ndarray, y_pred: np.ndarray) -> float:
    """
    Compute the mean squared error (MSE).

    Parameters
    ----------
    y_true : np.ndarray
        Ground truth target values of shape (n_samples,).
    y_pred : np.ndarray
        Predicted target values of shape (n_samples,).

    Returns
    -------
    float
        Mean squared error.
    """
    y_true_arr, y_pred_arr = _validate_regression_inputs(y_true=y_true, y_pred=y_pred)
    residuals = y_true_arr - y_pred_arr
    return float(np.mean(residuals**2))

r2_score 

r2_score(y_true, y_pred)

Compute the coefficient of determination (R² score).

Parameters:

Name Type Description Default
y_true ndarray

Ground truth target values of shape (n_samples,).

 required
y_pred ndarray

Predicted target values of shape (n_samples,).

 required

Returns:

Type Description
float

R² score.
Notes

If y_true is constant, returns 1.0 for perfect predictions and 0.0 otherwise.

Source code in glassbox/metrics/regression.py 
 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
def r2_score(y_true: np.ndarray, y_pred: np.ndarray) -> float:
    """
    Compute the coefficient of determination (R² score).

    Parameters
    ----------
    y_true : np.ndarray
        Ground truth target values of shape (n_samples,).
    y_pred : np.ndarray
        Predicted target values of shape (n_samples,).

    Returns
    -------
    float
        R² score.

    Notes
    -----
    If `y_true` is constant, returns 1.0 for perfect predictions and 0.0 otherwise.
    """
    y_true_arr, y_pred_arr = _validate_regression_inputs(y_true=y_true, y_pred=y_pred)

    ss_res = np.sum((y_true_arr - y_pred_arr) ** 2)
    y_mean = np.mean(y_true_arr)
    ss_tot = np.sum((y_true_arr - y_mean) ** 2)

    if ss_tot == 0.0:
        return 1.0 if ss_res == 0.0 else 0.0

    return float(1.0 - (ss_res / ss_tot))