Matthews correlation coefficient

The problem, from first principles

F1 ignores true negatives entirely, so it can look good on a model that is mostly wrong about the negatives. The Matthews correlation coefficient (MCC) uses all four cells of the confusion matrix. It is the correlation between predictions and truth: +1 means perfect agreement, 0 means no better than chance, and -1 means every prediction is inverted.

Implement matthews_corrcoef(y_true, y_pred, positive=1) for binary labels.

MCC = (TP · TN − FP · FN) / sqrt((TP + FP)(TP + FN)(TN + FP)(TN + FN))

If any factor in the denominator is zero, return 0.0.

Constraints

• Inputs must have the same length, otherwise raise ValueError.
• Any label other than positive counts as negative.

Hints

Open one at a time. Each gives away a little more than the last.