Regression Evaluation Metrics — ML Technical Interview Questions & Answers

MAE, MSE, RMSE and R² explained.

Regression Evaluation Metrics

  • Simple meaning: For number predictions, you measure how far guesses are from true marks. MAE = average absolute miss (easy to say in marks). MSE = average squared miss. RMSE = sqrt(MSE), same units as marks. R² = how much better than guessing the mean.
  • Picture: True 70, guess 80 → miss 10. MAE treats that as 10. RMSE punishes a miss of 30 harder than two misses of 10.
  • R² = 1 is perfect. R² = 0 is “no better than always guess average marks”. Negative R² means worse than that dummy.
  • Trap: quoting only train RMSE. Trap: comparing RMSE across datasets with different mark scales without saying so.

Technical Interview Tip: MAE = average miss. RMSE punishes big misses. R² vs guessing the mean.

Code Examples for Regression Evaluation Metrics

Regression Evaluation Metrics — working example

y_true = [1, 0, 1, 1, 0, 1]
y_pred = [1, 0, 0, 1, 0, 1]
correct = sum(t == p for t, p in zip(y_true, y_pred))
accuracy = correct / len(y_true)
print(correct, accuracy)
# Accuracy alone misleads on imbalanced classes — check precision/recall too.

Regression Evaluation Metrics — accuracy = correct / total. For imbalanced data, also report precision/recall/F1.

How to Explain Regression Evaluation Metrics to an Interviewer

MAE vs RMSE?

RMSE squares misses Big errors hurt RMSE more

R² = 0 means?

No better than predicting the average y