CBSE Class 10 AI Unit 3 MCQs with Answers & Explanations (2026–27) | Evaluating Models
Class 10 · Artificial Intelligence
Prepare for your CBSE Class 10 Artificial Intelligence (Code 417) examination with these 50 important Multiple Choice Questions (MCQs) from Unit 3: Evaluating Models. Every question includes the correct answer with a detailed explanation, helping you understand model evaluation techniques and score better in CBSE competency-based examinations.
Topics Covered
- Model Evaluation
- Importance of Model Evaluation
- Train-Test Split
- Training Dataset & Testing Dataset
- Error and Accuracy
- Classification Metrics
- Confusion Matrix
- True Positive (TP), True Negative (TN)
- False Positive (FP), False Negative (FN)
- Precision, Recall & F1 Score
- Ethics in Model Evaluation
- Competency-Based MCQs
CBSE Exam Tip: Learn the formulas for Accuracy, Precision, Recall, and F1 Score, understand the confusion matrix thoroughly, and know when each metric should be used. Numerical and assertion-reason questions are frequently asked from this unit.
Multiple Choice Questions (MCQs)
Q1. What is "Model Evaluation" in the context of Artificial Intelligence?
A. The process of collecting data from various sources
B. The process of using metrics to understand a machine learning model's performance
C. The final stage of deploying a model to a user interface
D. The process of choosing between rule-based and learning-based approaches
Show Answer & Explanation
Correct Answer: B. The process of using metrics to understand a machine learning model's performance
Explanation: Model Evaluation is the process of measuring how well a machine learning model performs by using different evaluation metrics. It helps developers identify the best model and estimate how accurately it will perform on new, unseen data.
Why other options are incorrect:
- A. Data collection is part of Data Acquisition, not Model Evaluation.
- C. Deployment happens after the model has been evaluated.
- D. Choosing between AI approaches is part of modelling, not evaluation.
Q2. Model Evaluation is compared to which real-life example in the CBSE handbook?
A. A supermarket trolley
B. A satellite launch pad
C. A school report card
D. A high-speed computer processor
Show Answer & Explanation
Correct Answer: C. A school report card
Explanation: Just as a school report card uses grades, percentages, and ranks to evaluate a student's performance, Model Evaluation uses various metrics to measure an AI model's performance and identify areas for improvement.
Why other options are incorrect:
- A. A supermarket trolley has no relation to performance evaluation.
- B. A launch pad is not used as an evaluation example in the syllabus.
- D. A processor performs computations but does not evaluate performance.
Q3. Why is Model Evaluation considered an important stage of AI development?
A. It helps find the best model and predicts future performance.
B. It eliminates the need for training data.
C. It guarantees 100% accuracy.
D. It replaces the modelling stage.
Show Answer & Explanation
Correct Answer: A. It helps find the best model and predicts future performance.
Explanation: Model Evaluation helps determine whether the selected model is suitable for solving the problem and estimates how well it will perform on unseen data.
Why other options are incorrect:
- B. Training data is essential for supervised learning.
- C. No AI model can guarantee perfect accuracy in all situations.
- D. Evaluation complements modelling; it does not replace it.
Q4. What helps an AI model improve its performance during Model Evaluation?
A. Unlabeled datasets only
B. Constructive feedback from evaluation metrics
C. Randomly generated rules
D. Higher screen resolution
Show Answer & Explanation
Correct Answer: B. Constructive feedback from evaluation metrics
Explanation: Evaluation metrics provide feedback about the model's strengths and weaknesses. Developers use this feedback to improve the model until satisfactory performance is achieved.
Why other options are incorrect:
- A. Unlabeled data alone cannot evaluate a supervised model.
- C. Random rules do not improve learning models.
- D. Screen resolution has no effect on model performance.
Q5. Which method is commonly used to evaluate a supervised learning model?
A. Feature Extraction
B. Principal Component Analysis
C. Train-Test Split
D. Data Acquisition
Show Answer & Explanation
Correct Answer: C. Train-Test Split
Explanation: In the Train-Test Split method, the dataset is divided into two parts: one for training the model and another for testing its performance.
Why other options are incorrect:
- A. Feature Extraction prepares data but does not evaluate models.
- B. Principal Component Analysis is a dimensionality reduction technique.
- D. Data Acquisition is the first stage of the AI Project Cycle.
Q6. What is the primary purpose of the Training Dataset?
A. To calculate the final accuracy
B. To make the model learn patterns from data
C. To display the final prediction
D. To store evaluation metrics
Show Answer & Explanation
Correct Answer: B. To make the model learn patterns from data
Explanation: The Training Dataset teaches the AI model by allowing it to identify relationships and patterns between input features and expected outputs.
Why other options are incorrect:
- A. Final accuracy is measured using the Testing Dataset.
- C. Predictions are generated after training.
- D. Evaluation metrics are calculated separately.
Q7. When is the Train-Test Split method most suitable?
A. When the dataset is extremely small
B. When a Rule-Based approach is used
C. When a sufficiently large dataset is available
D. Only for image segmentation problems
Show Answer & Explanation
Correct Answer: C. When a sufficiently large dataset is available
Explanation: A large dataset ensures that enough data is available for both training and independent testing, leading to reliable evaluation results.
Why other options are incorrect:
- A. Small datasets may not provide reliable evaluation.
- B. Rule-Based systems generally do not require Train-Test Split.
- D. Train-Test Split is used across many supervised learning problems.
Q8. After training, how is a machine learning model tested?
A. By giving it the same training dataset
B. By providing the testing dataset and comparing predictions with actual values
C. By manually changing its algorithm
D. By increasing the number of hidden layers
Show Answer & Explanation
Correct Answer: B. By providing the testing dataset and comparing predictions with actual values
Explanation: The trained model is tested using unseen data. Its predictions are compared with the actual labels to measure performance.
Why other options are incorrect:
- A. Testing on training data gives misleading results.
- C. Algorithms are not manually rewritten during testing.
- D. Adding hidden layers is part of model design, not testing.
Q9. The difference between the model's prediction and the actual result is called:
A. Precision
B. Accuracy
C. F1 Score
D. Error
Show Answer & Explanation
Correct Answer: D. Error
Explanation: Error measures how far a prediction is from the actual value or how frequently the model makes incorrect predictions.
Why other options are incorrect:
- A. Precision measures the correctness of positive predictions.
- B. Accuracy measures overall correctness.
- C. F1 Score combines Precision and Recall.
Q10. If a model correctly predicts whether a patient has a disease or not, the model has achieved good:
A. High Variance
B. Accuracy
C. False Positive Rate
D. Error Rate
Show Answer & Explanation
Correct Answer: B. Accuracy
Explanation: Accuracy measures the proportion of correct predictions made by the model out of all predictions. A higher accuracy generally indicates better overall performance.
Why other options are incorrect:
- A. High variance usually indicates overfitting.
- C. False Positive Rate measures one specific type of error.
- D. Error Rate represents incorrect predictions, not correct ones.
Q11. The performance of a model and its accuracy are related in which way?
A. They are inversely proportional.
B. They are not related.
C. They are directly proportional.
D. They change randomly.
Show Answer & Explanation
Correct Answer: C. They are directly proportional.
Explanation: As the performance of an AI model improves, the number of correct predictions increases. Therefore, better model performance generally leads to higher accuracy.
Why other options are incorrect:
- A. Better performance does not reduce accuracy.
- B. Performance and accuracy are closely related.
- D. Their relationship is not random.
Q12. What is the primary goal of an AI developer during Model Evaluation?
A. Maximize error and minimize accuracy
B. Minimize error and maximize accuracy
C. Increase the size of the dataset only
D. Increase the number of False Negatives
Show Answer & Explanation
Correct Answer: B. Minimize error and maximize accuracy
Explanation: During evaluation, developers try to reduce prediction errors while increasing the overall accuracy so that the model performs reliably on unseen data.
Why other options are incorrect:
- A. This is the opposite of the objective.
- C. A larger dataset alone does not guarantee better performance.
- D. Increasing False Negatives reduces model performance.
Q13. Why is relying only on Accuracy sometimes a problem in medical diagnosis?
A. Doctors do not use AI.
B. Accuracy does not consider the importance of different types of errors.
C. Accuracy can only be calculated for regression models.
D. Medical images are too large for evaluation.
Show Answer & Explanation
Correct Answer: B. Accuracy does not consider the importance of different types of errors.
Explanation: In healthcare, missing a sick patient (False Negative) can have serious consequences. Accuracy alone cannot show whether such critical mistakes are occurring.
Why other options are incorrect:
- A. AI is widely used in healthcare.
- C. Accuracy is commonly used for classification models.
- D. Image size has nothing to do with the usefulness of accuracy.
Q14. According to the handbook, the Error Rate is calculated as:
A. Predicted − Actual
B. 1 − Accuracy
C. TP + TN
D. Error ÷ Actual
Show Answer & Explanation
Correct Answer: D. Error ÷ Actual
Explanation: The handbook defines Error Rate as the ratio of the absolute error to the actual value, indicating how far the prediction is from the true value.
Why other options are incorrect:
- A. This calculates only the difference, not the error rate.
- B. Although commonly used elsewhere, this is not the handbook's stated formula.
- C. TP + TN represents correct predictions.
Q15. How is Accuracy Percentage calculated?
A. Accuracy ÷ 100
B. Accuracy × 100
C. Error × 100
D. (TP + FP) ÷ 100
Show Answer & Explanation
Correct Answer: B. Accuracy × 100
Explanation: To express Accuracy as a percentage, multiply the calculated accuracy value by 100.
Why other options are incorrect:
- A. Dividing by 100 decreases the value.
- C. This gives the error percentage, not accuracy.
- D. This is not the formula for accuracy percentage.
Q16. A Classification problem predicts:
A. Continuous numerical values
B. Specific class labels or categories
C. Pixel intensity values only
D. Mathematical weights
Show Answer & Explanation
Correct Answer: B. Specific class labels or categories
Explanation: Classification algorithms assign input data to predefined categories such as Spam/Not Spam or Pass/Fail.
Why other options are incorrect:
- A. Continuous values are predicted using Regression.
- C. Pixel values are image data, not classification outputs.
- D. Weights are parameters inside machine learning models.
Q17. Which of the following is NOT a popular evaluation metric for Classification models?
A. Confusion Matrix
B. Mean Absolute Error (MAE)
C. Precision
D. Recall
Show Answer & Explanation
Correct Answer: B. Mean Absolute Error (MAE)
Explanation: Confusion Matrix, Precision, Recall, Accuracy, and F1 Score are commonly used for evaluating classification models. Mean Absolute Error (MAE) is mainly used for regression problems.
Why other options are incorrect:
- A. It is a standard classification evaluation tool.
- C. Precision measures the correctness of positive predictions.
- D. Recall measures how many actual positive cases are identified.
Q18. In a Confusion Matrix, the Actual values are generally placed on the:
A. X-axis
B. Diagonal axis only
C. Y-axis
D. Z-axis
Show Answer & Explanation
Correct Answer: C. Y-axis
Explanation: In a standard Confusion Matrix, the Actual values are represented along the Y-axis, while the Predicted values are shown along the X-axis.
Why other options are incorrect:
- A. The X-axis represents Predicted values.
- B. The diagonal contains correct predictions, not actual values.
- D. A Confusion Matrix has no Z-axis.
Q19. When a model correctly predicts a positive class, the result is called:
A. True Negative (TN)
B. False Positive (FP)
C. True Positive (TP)
D. False Negative (FN)
Show Answer & Explanation
Correct Answer: C. True Positive (TP)
Explanation: A True Positive occurs when the model correctly predicts an instance that actually belongs to the positive class.
Why other options are incorrect:
- A. TN is a correct negative prediction.
- B. FP is an incorrect positive prediction.
- D. FN is an incorrect negative prediction.
Q20. If you predict that Germany will win the World Cup but they actually lose, this is an example of:
A. True Positive
B. False Positive
C. True Negative
D. False Negative
Show Answer & Explanation
Correct Answer: B. False Positive
Explanation: A False Positive occurs when the model predicts a positive outcome, but the actual result is negative. In this example, predicting Germany as the winner when they lose is a False Positive.
Why other options are incorrect:
- A. The prediction is not correct.
- C. No negative prediction was made.
- D. A False Negative means predicting negative when the actual outcome is positive.
Q21. If a person is actually sick (Positive) but the AI predicts they are healthy (Negative), this outcome is called:
A. True Positive (TP)
B. True Negative (TN)
C. False Positive (FP)
D. False Negative (FN)
Show Answer & Explanation
Correct Answer: D. False Negative (FN)
Explanation: A False Negative occurs when the actual class is positive, but the model incorrectly predicts it as negative. In medical diagnosis, this is one of the most dangerous types of errors.
Why other options are incorrect:
- A. True Positive means the positive case is correctly identified.
- B. True Negative means a negative case is correctly identified.
- C. False Positive means a negative case is wrongly predicted as positive.
Q22. Correctly predicting that a healthy person does not have a disease is an example of:
A. True Positive (TP)
B. True Negative (TN)
C. False Positive (FP)
D. False Negative (FN)
Show Answer & Explanation
Correct Answer: B. True Negative (TN)
Explanation: A True Negative occurs when the actual class is negative and the model also predicts it as negative.
Why other options are incorrect:
- A. TP refers to correctly identifying a positive case.
- C. FP predicts a healthy person as sick.
- D. FN predicts a sick person as healthy.
Q23. What is the formula for Classification Accuracy?
A. TP / (TP + FP)
B. (TP + TN) / (TP + TN + FP + FN)
C. TN / (TN + FN)
D. (TP + FN) / Total
Show Answer & Explanation
Correct Answer: B. (TP + TN) / (TP + TN + FP + FN)
Explanation: Classification Accuracy is calculated by dividing the total number of correct predictions (True Positives + True Negatives) by the total number of predictions.
Why other options are incorrect:
- A. This is the formula for Precision.
- C. This is not an evaluation metric formula.
- D. This does not calculate Accuracy.
Q24. A model that predicts "Yes" for every input regardless of the data is considered:
A. Highly reliable
B. A faulty model
C. A Rule-Based model
D. An Unsupervised Learning model
Show Answer & Explanation
Correct Answer: B. A faulty model
Explanation: A model that always predicts the same output ignores the input data and cannot make meaningful decisions. Therefore, it is considered a faulty model.
Why other options are incorrect:
- A. Such a model is unreliable.
- C. Rule-Based systems follow predefined rules, not a single constant output.
- D. This has nothing to do with Unsupervised Learning.
Q25. Accuracy is considered a reliable evaluation metric mainly when:
A. The dataset is highly unbalanced.
B. Both classes have nearly equal observations (balanced dataset).
C. Reinforcement Learning is used.
D. Only False Positives matter.
Show Answer & Explanation
Correct Answer: B. Both classes have nearly equal observations (balanced dataset).
Explanation: Accuracy gives meaningful results when the dataset is balanced. For highly unbalanced datasets, other metrics like Precision, Recall, and F1 Score are preferred.
Why other options are incorrect:
- A. Accuracy can be misleading for unbalanced datasets.
- C. Reinforcement Learning does not determine the suitability of Accuracy.
- D. Accuracy considers all predictions, not only False Positives.
Q26. Which evaluation metrics are recommended for unbalanced datasets?
A. Pixel Resolution and Brightness
B. Precision, Recall, and F1 Score
C. Mean and Median
D. Regression and Association
Show Answer & Explanation
Correct Answer: B. Precision, Recall, and F1 Score
Explanation: Precision, Recall, and F1 Score provide a much better understanding of model performance when one class is significantly larger than the other.
Why other options are incorrect:
- A. These are image characteristics, not evaluation metrics.
- C. Mean and Median are statistical measures.
- D. Regression and Association are learning techniques, not evaluation metrics.
Q27. Precision is defined as the ratio of correctly predicted positive cases to:
A. Total actual positive cases
B. Total correct predictions
C. Total predicted positive cases
D. Total True Negative cases
Show Answer & Explanation
Correct Answer: C. Total predicted positive cases
Explanation: Precision = TP / (TP + FP). It measures how many of the cases predicted as positive are actually positive.
Why other options are incorrect:
- A. This relates to Recall.
- B. Overall correct predictions are used for Accuracy.
- D. True Negatives are not included in the Precision formula.
Q28. Precision is especially important when reducing which type of error?
A. False Negatives (FN)
B. True Positives (TP)
C. False Positives (FP)
D. True Negatives (TN)
Show Answer & Explanation
Correct Answer: C. False Positives (FP)
Explanation: Precision becomes important when False Positives are costly or dangerous. A high Precision value indicates that most positive predictions are correct.
Why other options are incorrect:
- A. False Negatives are mainly handled using Recall.
- B. True Positives should be maximized, not reduced.
- D. True Negatives are not the focus of Precision.
Q29. Which of the following situations requires high Precision?
A. Predicting Covid-19 infections
B. Detecting fraudulent credit card transactions
C. Launching a satellite based on weather conditions
D. Diagnosing cancer
Show Answer & Explanation
Correct Answer: C. Launching a satellite based on weather conditions
Explanation: A False Positive (predicting safe weather when it is actually unsafe) could result in a failed satellite launch. Therefore, high Precision is essential.
Why other options are incorrect:
- A. Covid-19 detection generally prioritizes Recall.
- B. Fraud detection usually focuses on minimizing missed frauds.
- D. Cancer diagnosis also emphasizes Recall.
Q30. Recall measures how well a model identifies:
A. All possible outputs
B. True Positives out of all actual positive cases
C. False Positives out of predicted positives
D. True Negatives out of actual negatives
Show Answer & Explanation
Correct Answer: B. True Positives out of all actual positive cases
Explanation: Recall = TP / (TP + FN). It measures the model's ability to correctly identify all actual positive cases.
Why other options are incorrect:
- A. Recall focuses only on positive cases.
- C. This describes Precision.
- D. True Negatives are not part of the Recall formula.
Q31. What is another name for Recall?
A. Accuracy Rate
B. Sensitivity
C. Specificity
D. Error Rate
Show Answer & Explanation
Correct Answer: B. Sensitivity
Explanation: Recall is also known as Sensitivity or the True Positive Rate. It measures how effectively a model identifies actual positive cases.
Why other options are incorrect:
- A. Accuracy measures overall correct predictions.
- C. Specificity measures the correct identification of negative cases.
- D. Error Rate measures incorrect predictions.
Q32. In which situation is Recall more important than Precision?
A. Marking a genuine email as spam
B. Predicting a Covid-19 infected person as "non-infected"
C. Predicting a sunny day for a picnic
D. Choosing a website theme
Show Answer & Explanation
Correct Answer: B. Predicting a Covid-19 infected person as "non-infected"
Explanation: Missing an infected patient (False Negative) can lead to delayed treatment and further spread of the disease. Therefore, Recall is more important in medical diagnosis.
Why other options are incorrect:
- A. This relates more to Precision.
- C. It is not a critical healthcare application.
- D. Website themes are unrelated to model evaluation.
Q33. The F1 Score combines which two evaluation metrics?
A. Accuracy and Error Rate
B. True Positive and True Negative
C. Precision and Recall
D. Bias and Transparency
Show Answer & Explanation
Correct Answer: C. Precision and Recall
Explanation: The F1 Score is the harmonic mean of Precision and Recall. It provides a balanced measure when both False Positives and False Negatives are important.
Why other options are incorrect:
- A. Accuracy and Error Rate are separate evaluation metrics.
- B. TP and TN are values in a Confusion Matrix.
- D. Bias and Transparency are ethical principles, not evaluation metrics.
Q34. When should an AI developer prefer using the F1 Score?
A. When the dataset is perfectly balanced
B. When only True Negatives matter
C. When both False Positives and False Negatives are important
D. Only while developing Rule-Based systems
Show Answer & Explanation
Correct Answer: C. When both False Positives and False Negatives are important
Explanation: F1 Score is used when developers cannot prioritize either Precision or Recall alone and require a balanced evaluation of both.
Why other options are incorrect:
- A. Balanced datasets do not specifically require F1 Score.
- B. F1 Score does not focus on True Negatives.
- D. Rule-Based systems do not use F1 Score for evaluation.
Q35. What is the mathematical formula for the F1 Score?
A. Precision + Recall
B. (Precision × Recall) ÷ 2
C. 2 × (Precision × Recall) ÷ (Precision + Recall)
D. (TP + TN) ÷ Total
Show Answer & Explanation
Correct Answer: C. 2 × (Precision × Recall) ÷ (Precision + Recall)
Explanation: The F1 Score is calculated using the harmonic mean of Precision and Recall, giving equal importance to both metrics.
Why other options are incorrect:
- A. Simply adding the values does not calculate F1 Score.
- B. This is the arithmetic mean, not the harmonic mean.
- D. This is the formula for Accuracy.
Q36. In fraud detection, why is Recall usually given higher importance?
A. Because verifying a genuine transaction is acceptable, but missing a fraudulent one is risky.
B. Because legitimate transactions are rare.
C. Because banks only calculate Accuracy.
D. Because False Positives are more dangerous than False Negatives.
Show Answer & Explanation
Correct Answer: A. Because verifying a genuine transaction is acceptable, but missing a fraudulent one is risky.
Explanation: Missing a fraudulent transaction (False Negative) may result in financial loss, whereas checking a genuine transaction again (False Positive) causes only minor inconvenience.
Why other options are incorrect:
- B. Legitimate transactions are generally more common.
- C. Banks use multiple evaluation metrics.
- D. In fraud detection, False Negatives are usually more harmful.
Q37. Which ethical principle ensures that AI evaluation does not unfairly discriminate against any group?
A. Accountability
B. Transparency
C. Accuracy
D. Bias
Show Answer & Explanation
Correct Answer: D. Bias
Explanation: During Model Evaluation, developers must identify and reduce bias to ensure that AI systems treat all individuals fairly and do not discriminate.
Why other options are incorrect:
- A. Accountability means taking responsibility for AI decisions.
- B. Transparency refers to openness about AI processes.
- C. Accuracy measures model performance but does not ensure fairness.
Q38. What does Transparency mean in Model Evaluation?
A. Hiding evaluation methods from users
B. Clearly explaining how evaluation metrics work and how results are obtained
C. Using transparent images only
D. Automating evaluation without human involvement
Show Answer & Explanation
Correct Answer: B. Clearly explaining how evaluation metrics work and how results are obtained
Explanation: Transparency ensures that developers and stakeholders understand how an AI model is evaluated and how its decisions are measured.
Why other options are incorrect:
- A. Transparency requires openness, not secrecy.
- C. Image transparency is unrelated to AI ethics.
- D. Transparency involves human understanding, not complete automation.
Q39. Taking responsibility for the evaluation metrics and methodology used in AI is known as:
A. Accountability
B. Beneficence
C. Non-maleficence
D. Data Exploration
Show Answer & Explanation
Correct Answer: A. Accountability
Explanation: Accountability means AI developers and organizations are responsible for the methods, decisions, and outcomes of their AI systems.
Why other options are incorrect:
- B. Beneficence means working for the benefit of society.
- C. Non-maleficence means avoiding harm.
- D. Data Exploration is another stage of the AI Project Cycle.
Q40. A model has TP = 150, FP = 50, TN = 750, and FN = 50. What is its Accuracy?
A. 80%
B. 95%
C. 90%
D. 75%
Show Answer & Explanation
Correct Answer: C. 90%
Explanation:
Accuracy = (TP + TN) / (TP + TN + FP + FN)
= (150 + 750) / (150 + 750 + 50 + 50)
= 900 / 1000 = 0.90 = 90%
Why other options are incorrect:
- A. Incorrect calculation.
- B. Overestimates the model's performance.
- D. Does not match the Accuracy formula.
Q41. Using the values TP = 150, FP = 50, TN = 750, and FN = 50, what is the Precision?
A. 0.75
B. 0.85
C. 0.90
D. 0.60
Show Answer & Explanation
Correct Answer: A. 0.75
Explanation:
Precision = TP / (TP + FP)
= 150 / (150 + 50)
= 150 / 200 = 0.75
Precision tells us how many of the predicted positive cases are actually positive.
Why other options are incorrect:
- B. Incorrect calculation.
- C. Overestimates the Precision.
- D. Does not satisfy the Precision formula.
Q42. Using the values TP = 150, FP = 50, TN = 750, and FN = 50, what is the Recall?
A. 0.90
B. 0.75
C. 0.80
D. 0.50
Show Answer & Explanation
Correct Answer: B. 0.75
Explanation:
Recall = TP / (TP + FN)
= 150 / (150 + 50)
= 150 / 200 = 0.75
Recall measures how many of the actual positive cases were correctly identified by the model.
Why other options are incorrect:
- A. Incorrect calculation.
- C. Does not match the Recall formula.
- D. Underestimates the Recall value.
Q43. If a student answers 90 out of 100 questions correctly, what is the Error Rate?
A. 90%
B. 10%
C. 5%
D. 0.1%
Show Answer & Explanation
Correct Answer: B. 10%
Explanation:
The student answered 10 questions incorrectly.
Error Rate = Incorrect Answers / Total Questions
= 10 / 100 = 10%
Why other options are incorrect:
- A. This represents the Accuracy.
- C. Incorrect calculation.
- D. Much smaller than the actual Error Rate.
Q44. What can happen if an AI model is deployed without proper evaluation?
A. It automatically becomes more accurate over time.
B. It may perform poorly on new data and produce unreliable predictions.
C. The model will refuse to execute.
D. It automatically changes into an Unsupervised Learning model.
Show Answer & Explanation
Correct Answer: B. It may perform poorly on new data and produce unreliable predictions.
Explanation: Without evaluation, developers cannot determine whether the model generalizes well to unseen data. This can result in inaccurate and unreliable predictions in real-world applications.
Why other options are incorrect:
- A. AI models do not automatically improve after deployment.
- C. Deployment does not stop the model from running.
- D. Learning approaches do not change automatically.
Q45. Which of the following is an example of a Binary Classification problem?
A. Predicting the price of a house
B. Predicting whether a loan applicant will "Default" or "Non-default"
C. Grouping songs according to mood
D. Translating English into French
Show Answer & Explanation
Correct Answer: B. Predicting whether a loan applicant will "Default" or "Non-default"
Explanation: Binary Classification problems have only two possible output classes. Here, the model predicts either Default or Non-default.
Why other options are incorrect:
- A. Predicting house prices is a Regression problem.
- C. Grouping songs is a Clustering problem.
- D. Language translation is a Natural Language Processing task.
Q46. Assertion (A): Accuracy is not always the best evaluation metric.
Reason (R): In an unbalanced dataset, a faulty model may still achieve high Accuracy.
A. Both A and R are true, and R is the correct explanation of A.
B. A is true, but R is false.
C. Both A and R are true, but R is not the correct explanation of A.
D. A is false, but R is true.
Show Answer & Explanation
Correct Answer: A. Both A and R are true, and R is the correct explanation of A.
Explanation: In highly unbalanced datasets, a model may correctly predict only the majority class and still obtain very high Accuracy. Therefore, metrics like Precision, Recall, and F1 Score become more meaningful.
Why other options are incorrect:
- B. The Reason is also true.
- C. The Reason correctly explains the Assertion.
- D. The Assertion is true.
Q47. A medical diagnostic test correctly identifies 90 out of 100 sick people but wrongly identifies 50 healthy people as sick. What is its Precision?
A. 90%
B. 64%
C. 50%
D. 80%
Show Answer & Explanation
Correct Answer: B. 64%
Explanation:
Precision = TP / (TP + FP)
= 90 / (90 + 50)
= 90 / 140 ≈ 0.64 = 64%
Why other options are incorrect:
- A. This is not obtained using the Precision formula.
- C. Incorrect calculation.
- D. Does not match the calculated Precision.
Q48. In the "Bob and Billy" concert example, the entry fee was ₹500. Bob brought ₹300. What was Bob's Absolute Error?
A. ₹500
B. ₹300
C. ₹200
D. ₹800
Show Answer & Explanation
Correct Answer: C. ₹200
Explanation:
Absolute Error = |Actual Value − Predicted Value|
= |₹500 − ₹300|
= ₹200
Why other options are incorrect:
- A. This is the actual amount, not the error.
- B. This is the amount Bob brought.
- D. Incorrect calculation.
Q49. The sum of all values in a Confusion Matrix represents:
A. The total number of correct predictions
B. The total number of instances in the testing dataset
C. The total prediction error
D. The total number of features
Show Answer & Explanation
Correct Answer: B. The total number of instances in the testing dataset
Explanation: The total number of test samples is calculated as:
TP + TN + FP + FN
This equals the total number of instances evaluated by the model.
Why other options are incorrect:
- A. Correct predictions include only TP + TN.
- C. The Confusion Matrix does not directly represent total error.
- D. Features are unrelated to the Confusion Matrix totals.
Q50. Which of the following is NOT a stage of the AI Project Cycle that comes before Model Evaluation?
A. Modelling
B. Data Exploration
C. Deployment
D. Problem Scoping
Show Answer & Explanation
Correct Answer: C. Deployment
Explanation: Deployment takes place after Model Evaluation. Before Evaluation, the AI Project Cycle includes Problem Scoping, Data Acquisition, Data Exploration, and Modelling.
Why other options are incorrect:
- A. Modelling is completed before Evaluation.
- B. Data Exploration is an earlier stage.
- D. Problem Scoping is the first stage of the AI Project Cycle.
? Quick Revision
- Model Evaluation measures how well an AI model performs using evaluation metrics.
- Train-Test Split divides data into a Training Dataset (for learning) and a Testing Dataset (for evaluation).
- Accuracy measures the overall percentage of correct predictions.
- Error is the difference between the predicted value and the actual value.
- Confusion Matrix consists of TP, TN, FP, and FN.
- Precision = TP / (TP + FP) — important when False Positives should be minimized.
- Recall = TP / (TP + FN) — important when False Negatives should be minimized.
- F1 Score is the harmonic mean of Precision and Recall and is useful when both metrics are equally important.
- Accuracy alone is not reliable for highly unbalanced datasets.
- Ethical Model Evaluation should ensure Fairness (Bias), Transparency, and Accountability.
? Exam Tip
For the CBSE Class 10 Artificial Intelligence (Code 417) examination, thoroughly understand the Confusion Matrix (TP, TN, FP, FN), memorize the formulas for Accuracy, Precision, Recall, and F1 Score, and know when each metric should be used. Practice numerical questions, assertion-reason questions, and competency-based case studies, as these are commonly asked in CBSE examinations.