MCQs / Computer Science / NSCT: AI, Machine Learning & Data Analytics
NSCT: AI, Machine Learning & Data Analytics MCQs — Computer Science 127 multiple-choice questions on NSCT: AI, Machine Learning & Data Analytics from Computer Science, each with the correct answer and, where available, a teacher's explanation. Read through them to revise the chapter, then practise the same questions one by one to test yourself.
1. When the statistical properties of live input data change after deployment, it is called: Overfitting during training Data drift Tokenization Dropout Answer: B) Data drift
Explanation: Models need monitoring and retraining. Lesson 8.5 · AI, Machine Learning & Data Analytics
Discuss this question
2. Standardization (z-score) transforms a feature to have: Integer values Mean 0 and standard deviation 1 Range 0 to 100 Only positive values Answer: B) Mean 0 and standard deviation 1
Explanation: z = (x − mean) ÷ std. Lesson 8.2 · AI, Machine Learning & Data Analytics
Discuss this question
3. Information about the test set accidentally used during training causes: Underfitting Data leakage and over-optimistic results Lower training accuracy Missing values Answer: B) Data leakage and over-optimistic results
Explanation: The model looks better than it really is. Lesson 8.3 · AI, Machine Learning & Data Analytics
Discuss this question
4. A search is guaranteed to find an optimal path when its heuristic is: Always zero only Random Larger than the true cost Admissible (never overestimates the true cost) Answer: D) Admissible (never overestimates the true cost)
Explanation: f(n) = g(n) + h(n). Lesson 8.1 · AI, Machine Learning & Data Analytics
Discuss this question
5. A feature ranges from 50 to 100. What is the min-max normalized value of 75? 1.5 0.6 0.5 0.75 Answer: C) 0.5
Explanation: (x − min) ÷ (max − min) = (75 − 50) ÷ (100 − 50) = 0.5. Lesson 8.2 · AI, Machine Learning & Data Analytics
Discuss this question
6. A slightly altered image that fools a classifier into a wrong prediction is an: Outlier in EDA Augmented sample for training only Adversarial example Embedding Answer: C) Adversarial example
Explanation: It is an AI security risk. Lesson 8.5 · AI, Machine Learning & Data Analytics
Discuss this question
7. Reducing "running", "ran" and "runs" to the base form "run" using vocabulary and grammar is: Lemmatization Tokenization Padding One-hot encoding Answer: A) Lemmatization
Explanation: Stemming just chops endings. Lesson 8.5 · AI, Machine Learning & Data Analytics
Discuss this question
8. PCA reduces 100 features to 10 principal components. How many features does each sample have after PCA? 11 90 100 10 Answer: D) 10
Explanation: Each sample is projected onto the chosen components. Lesson 8.2 · AI, Machine Learning & Data Analytics
Discuss this question
9. Which statistic is least affected by outliers? Median Maximum Range Mean Answer: A) Median
Explanation: The median depends only on the middle value. Lesson 8.2 · AI, Machine Learning & Data Analytics
Discuss this question
10. A classifier's confusion matrix is: TP = 25, FP = 25, FN = 25, TN = 25. What is its recall? 52% 49% 51% 50% Answer: D) 50%
Explanation: Recall = TP ÷ (TP + FN) = 25 ÷ 50 = 50%. Lesson 8.3 · AI, Machine Learning & Data Analytics
Discuss this question
11. To see the relationship between two numeric variables, the best chart is a: Scatter plot Histogram of one variable Pie chart Bar chart of categories Answer: A) Scatter plot
Explanation: Each point is one observation. Lesson 8.2 · AI, Machine Learning & Data Analytics
Discuss this question
12. For cancer screening, missing a sick patient is very costly. Which metric matters most? Specificity only Recall Precision Accuracy Answer: B) Recall
Explanation: Recall = TP ÷ (TP + FN). Lesson 8.3 · AI, Machine Learning & Data Analytics
Discuss this question
13. Actual values [3, 5, 2, 7] and predictions [4, 5, 3, 5]. What is the Mean Absolute Error (MAE)? 1.22 1.5 0 1 Answer: D) 1
Explanation: Errors [-1, 0, -1, 2]; mean of absolute values = 1. Lesson 8.3 · AI, Machine Learning & Data Analytics
Discuss this question
14. Which technique helps reduce overfitting? Testing on training data Training longer forever Adding more irrelevant features Regularization (L1/L2) Answer: D) Regularization (L1/L2)
Explanation: It penalises large weights. Lesson 8.3 · AI, Machine Learning & Data Analytics
Discuss this question
15. An expert system consists mainly of a knowledge base and an: Inference engine Operating system Compiler Activation function Answer: A) Inference engine
Explanation: The inference engine applies rules to facts. Lesson 8.1 · AI, Machine Learning & Data Analytics
Discuss this question
16. Apache Spark is faster than classic Hadoop MapReduce mainly because it: Uses no cluster Is single-threaded Stores data only on tape Processes data in memory Answer: D) Processes data in memory
Explanation: It reduces disk I/O. Lesson 8.5 · AI, Machine Learning & Data Analytics
Discuss this question
17. In k-NN with k = 3, the classes of the 3 nearest neighbours are: A, B, A. What class is predicted? A tie – no prediction B A The class of the farthest neighbour Answer: C) A
Explanation: Majority vote: A (2 of 3). Lesson 8.3 · AI, Machine Learning & Data Analytics
Discuss this question
18. Using a small labelled set and a large unlabelled set together is: Reinforcement learning Pure supervised learning Semi-supervised learning Rule-based AI Answer: C) Semi-supervised learning
Explanation: Useful when labelling is expensive. Lesson 8.4 · AI, Machine Learning & Data Analytics
Discuss this question
19. In k-NN with k = 5, the classes of the 5 nearest neighbours are: Spam, Ham, Ham, Spam, Ham. What class is predicted? Ham Spam A tie – no prediction The class of the farthest neighbour Answer: A) Ham
Explanation: Majority vote: Ham (3 of 5). Lesson 8.3 · AI, Machine Learning & Data Analytics
Discuss this question
20. One complete pass of the whole training dataset through the network is an: Layer Epoch Batch Iteration of one sample only Answer: B) Epoch
Explanation: Training usually needs many epochs. Lesson 8.5 · AI, Machine Learning & Data Analytics
Discuss this question
21. A classifier's confusion matrix is: TP = 40, FP = 10, FN = 5, TN = 45. What is its F1-score (to 2 decimal places)? 0.84 0.71 0.85 0.89 Answer: A) 0.84
Explanation: F1 = 2PR ÷ (P + R) = 2 × 0.80 × 0.89 ÷ (0.80 + 0.89) = 0.84. Lesson 8.3 · AI, Machine Learning & Data Analytics
Discuss this question
22. When an LLM confidently generates false information, it is called: Hallucination Clustering Tokenization Overfitting only Answer: A) Hallucination
Explanation: Outputs must be verified. Lesson 8.5 · AI, Machine Learning & Data Analytics
Discuss this question
23. In k-means clustering, k is: The number of clusters chosen in advance The learning rate The number of features The number of neighbours Answer: A) The number of clusters chosen in advance
Explanation: The elbow method helps choose k. Lesson 8.4 · AI, Machine Learning & Data Analytics
Discuss this question
24. For multi-class classification, the output layer usually uses: Softmax ReLU No activation Sigmoid for each class only in all cases Answer: A) Softmax
Explanation: Softmax gives probabilities that sum to 1. Lesson 8.5 · AI, Machine Learning & Data Analytics
Discuss this question
25. "Satisfaction: low, medium, high" is which type of data? Nominal categorical Continuous numeric Ordinal categorical Unstructured Answer: C) Ordinal categorical
Explanation: The categories have a natural order. Lesson 8.2 · AI, Machine Learning & Data Analytics
Discuss this question
Page 1 of 6 Next
Practise NSCT: AI, Machine Learning & Data Analytics MCQs
More Computer Science chapters