ML Metrics¶

These functions compute standard machine learning evaluation metrics.

accuracy¶

accuracy(y_true, y_pred)

Computes classification accuracy: the proportion of correct predictions.

Parameters:

• y_true - Array of true labels
• y_pred - Array of predicted labels

Returns: A number between 0.0 and 1.0.

let acc = accuracy([1, 0, 1, 1], [1, 0, 0, 1]);
// acc = 0.75

f1_score¶

f1_score(y_true, y_pred)

Computes the F1 score, the harmonic mean of precision and recall.

Parameters:

• y_true - Array of true labels
• y_pred - Array of predicted labels

Returns: A number between 0.0 and 1.0.

let f1 = f1_score([1, 0, 1, 1], [1, 0, 0, 1]);

precision¶

precision(y_true, y_pred)

Computes precision: the proportion of positive predictions that are correct.

Parameters:

• y_true - Array of true labels
• y_pred - Array of predicted labels

Returns: A number between 0.0 and 1.0.

let prec = precision([1, 0, 1, 1], [1, 0, 0, 1]);

recall¶

recall(y_true, y_pred)

Computes recall: the proportion of actual positives that were correctly predicted.

Parameters:

• y_true - Array of true labels
• y_pred - Array of predicted labels

Returns: A number between 0.0 and 1.0.

let rec = recall([1, 0, 1, 1], [1, 0, 0, 1]);

confusion_matrix¶

confusion_matrix(y_true, y_pred)

Computes a multi-class confusion matrix.

Parameters:

• y_true - Array of true labels
• y_pred - Array of predicted labels

Returns: A 2D array where element [i][j] is the number of samples with true label i predicted as label j.

let cm = confusion_matrix(ground_truth, predictions);

roc_curve¶

roc_curve(y_true, y_pred)

Computes the Receiver Operating Characteristic (ROC) curve.

Parameters:

• y_true - Array of true binary labels
• y_pred - Array of predicted probabilities

Returns: A record with three fields:

• "fpr" - Array of false positive rates
• "tpr" - Array of true positive rates
• "thresholds" - Array of thresholds

let roc = roc_curve(labels, probabilities);

brier_score_model¶

brier_score_model(y_true, y_pred)

Computes the Brier score (mean squared error) for probability predictions over the entire dataset.

Parameters:

• y_true - Array of true binary labels (0 or 1)
• y_pred - Array of predicted probabilities

Returns: A single number. Lower is better; 0.0 is a perfect score.

let brier = brier_score_model(labels, probabilities);

brier_score_element¶

brier_score_element(y_true, y_pred)

Computes per-element Brier scores, returning a score for each prediction.

Parameters:

• y_true - Array of true binary labels (0 or 1)
• y_pred - Array of predicted probabilities

Returns: An array of per-element squared errors.

let element_scores = brier_score_element(labels, probabilities);