Implement classification_metrics(y_true, y_pred) — the confusion matrix and its derived metrics for binary classification, without sklearn.

A staple of MLE interviews, and what separates answers is how you handle division by zero.

Input

Two lists of the same length. Each element is 0 or 1.

y_true = [1, 1, 0, 1, 0]
y_pred = [1, 0, 0, 1, 1]

Confusion matrix

The positive class is 1.

TP = actual 1, predicted 1        FN = actual 1, predicted 0
FP = actual 0, predicted 1        TN = actual 0, predicted 0

Return

{"tp": 2, "tn": 1, "fp": 1, "fn": 1,
 "precision": 0.666667, "recall": 0.666667, "f1": 0.666667}

Metric definitions

precision = TP / (TP + FP)          ← denominator is "what we predicted positive"
recall    = TP / (TP + FN)          ← denominator is "what actually is positive"
f1        = 2·precision·recall / (precision + recall)

Non-negotiables

• Never divide by zero. When a denominator is zero, that metric is 0.0.
  • Nothing predicted positive → precision = 0.0
  • No actual positives → recall = 0.0
  • precision + recall == 0 → f1 = 0.0
• Round all three metrics to 6 decimal places (round(x, 6)).
• Empty input ([], []) gives four zero counts and three 0.0 metrics.

Implement classification_metrics(y_true, y_pred).

• The two lists always have the same length.
• The returned dictionary has exactly these seven keys: tp, tn, fp, fn, precision, recall, f1.
• Counts (tp and friends) are ints; metrics are floats.