Evaluating Models - Free MCQ Practice Test with solutions, Class 10 Artificial

Evaluation is the process of using metrics (like accuracy, precision, recall, F1) to understand how well a model performs, its strengths/weaknesses, and where it needs improvement.

The chapter emphasizes iterative improvement: build a model, evaluate with metrics, refine, and continue until desired performance is reached.

Train–test split separates data for learning (train) and for objective evaluation (test), estimating performance on new, unseen data.

The test set is used after training to gauge how well the model generalizes to data it never saw during training.

TP = correctly predicted positive; FN = actual positive predicted as negative (missed positive).

Error = 402 − 391 = 11; Error Rate = 11/402 ≈ 0.0274; Accuracy = 1 − 0.0274 = 0.9726 ≈ 97.3%.

Error = 500 − 465 = 35; Error Rate = 35/500 = 0.07; Accuracy = 1 − 0.07 = 0.93 → 93%.

Accuracy measures all correct predictions: (TP + TN) divided by all predictions.

Precision measures correctness among positive predictions: TP divided by all predicted positives (TP + FP).

Recall measures how many actual positives were captured: TP divided by actual positives (TP + FN).

F1 = 2PR/(P + R), the harmonic mean balancing precision and recall, especially useful when both matter.

Accuracy works best with balanced datasets and when FP and FN have comparable costs; otherwise use precision/recall/F1.

Accuracy = (TP + TN)/Total = (4 + 2)/10 = 6/10 = 60%.

Precision = TP/(TP + FP) = 50/60 = 0.8333 → 83.33%.

Recall = TP/(TP + FN) = 50/50 = 1.0 → 100%.

F1 = 2PR/(P+R) = 2×0.8333×1 / (1.8333) ≈ 0.9091 → 90.91%.

Total = 100 + 47 + 62 + 290 = 499; Accuracy = (TP + TN)/Total = 147/499 ≈ 0.2946 → 29.46%.

If the model predicts only the majority class, accuracy can look high while performance on the minority class is poor—hence consider precision/recall/F1.

Transparency ensures decision processes are explainable (e.g., explaining why a loan was denied), fostering trust.

Bias is unfairness/discrimination in outcomes—e.g., a system favoring one group (like only showing products to male users), leading to unfair results and poorer business outcomes.