1. What is the primary goal of extracting knowledge from big datasets?

Answer: b) Pattern recognition for decision-making

Explanation:

a) Data storage optimization – Not the main goal, though efficient storage is important.

b) Pattern recognition for decision-making – Correct. Data science aims to discover patterns to
guide strategic actions.

c) Hardware maintenance – Irrelevant to data analysis.

d) Network security – A separate domain, though it can use data science techniques.

e) Software licensing – Not related to extracting knowledge from data.

---

2. Which data type is characterized by a fixed schema?

Answer: c) Structured
Explanation:

a) Unstructured – No predefined format (e.g., text, video).

b) Semi-structured – Has some structure (e.g., XML, JSON) but not rigid.

c) Structured – Correct. Follows a fixed schema like relational databases.

d) Temporal – Relates to time-based data, not schema.

e) Geospatial – Data with location info; structure varies.

---

3. The data science process step that involves handling outliers occurs during:

Answer: b) Pre-processing
Explanation:

a) Data collection – Gathering raw data, not modifying it.

b) Pre-processing – Correct. Cleaning data, handling outliers and missing values.

c) Training – Applies models to cleaned data.

d) Deployment – Putting models into production.

e) Visualization – Shows data, doesn't modify it.

---

4. Which tool is specifically designed for distributed data processing?

Answer: b) Hadoop
Explanation:

a) Excel – For local spreadsheet tasks.

b) Hadoop – Correct. Distributed storage and processing of large datasets.

c) Photoshop – For image editing.

d) PowerPoint – For presentations.

e) WordPress – Web content management.

---

5. A box plot is most effective for displaying:

Answer: b) Data distributions

Explanation:

a) Time-series trends – Better shown with line charts.

b) Data distributions – Correct. Shows median, quartiles, and outliers.

c) Geographic patterns – Maps or GIS visualizations are used.

d) Text sentiment – Bar or word clouds are better.

e) Audio waveforms – Use waveform or spectrogram plots.

---

6. In probability theory, the Poisson distribution is typically used to model:

Answer: c) Rare event counts

Explanation:

a) Continuous measurements – Use normal or exponential distributions.

b) Binary outcomes – Use Bernoulli or Binomial.

c) Rare event counts – Correct. Like calls per hour or website hits.

d) Circular data – Modeled with von Mises distribution.

e) Categorical variables – Use Multinomial distribution.

7. Which phase of the data science process includes feature engineering?

Answer: b) Pre-processing
Explanation:

a) Data collection – Gathering data, no modifications.

b) Pre-processing – Correct. Includes feature creation/modification.

c) Model deployment – Using the model in production.

d) Business reporting – Presenting results, not transforming features.

e) Hardware setup – Infrastructure, not data handling.

---
8. The term "ETL" in data management refers to:

Answer: b) Extract, Transform, Load

Explanation:

a) Error Tracking Logs – Unrelated.

b) Extract, Transform, Load – Correct. Pipeline to prepare data for analysis.

c) Embedded Testing Layers – Not applicable.

d) Encryption Transfer Logic – Not a standard term.

e) External Tagging Libraries – Doesn't describe data pipelines.

---

9. Which is a key characteristic of unstructured data?

Answer: c) No predefined format

Explanation:

a) Fits relational databases – That's structured data.

b) Contains metadata tags – More common in semi-structured data.

c) No predefined format – Correct. Text, audio, video, etc.

d) Easy to query with SQL – SQL works with structured data.

e) Always numerical – Not true; can be text, images, etc.

---

10. In the CRISP-DM methodology, which phase follows business understanding?

Answer: e) Data understanding

Explanation:

a) Data preparation – Happens after understanding data.

b) Modeling – Later step.

c) Evaluation – Near the end.

d) Deployment – Final stage.

e) Data understanding – Correct. Second phase after defining business goals

21. In regression, multicollinearity refers to:

Answer: b) High correlation among independent variables

a) Outliers in data – Different issue.

b) Correct. It causes instability in coefficient estimates.

c) Non-linear relationships – Not relevant here.

d) Missing values – Separate problem.

e) Response variable variability – Not multicollinearity.

---

22. The term “dimensionality reduction” refers to:

Answer: c) Reducing number of features

a) Normalizing data – Scaling.

b) Reducing file size – Possible side effect.

c) Correct. It simplifies models and reduces overfitting.

d) Increasing model complexity – Opposite goal.

e) Reducing training time – A benefit, not the definition.

---
23. Which of the following is a supervised learning task?

Answer: a) Classification

a) Correct. Labeled data is used to assign categories.

b) Clustering – Unsupervised.

c) Dimensionality reduction – Often unsupervised.

d) Association rule mining – Unsupervised.

e) Topic modeling – Unsupervised.

---

24. What does overfitting refer to in machine learning?

Answer: c) Model fits training data too well

a) Too few features – May cause underfitting.

b) Low training accuracy – Actually sign of underfitting.

c) Correct. Overfitted models perform poorly on new data.

d) High generalization – This is desired, not a problem.

e) Limited data – Can contribute, but not definition.

---

25. Which data visualization is best for showing parts of a whole?

Answer: b) Pie chart

a) Scatter plot – Relationships.

b) Correct. Pie charts show proportions.

c) Histogram – Distribution.

d) Box plot – Quartiles and spread.

e) Line graph – Time trends.

---

26. Which is NOT a component of the data science Venn diagram (Drew Conway)?

Answer: e) Graphic design

a) Hacking skills

b) Math/stat knowledge

c) Substantive expertise

d) All of the above are included

e) Correct. Graphic design is not part of the core Venn.

---

27. Which database system is NoSQL?

Answer: d) MongoDB

a) MySQL – SQL.

b) PostgreSQL – SQL.

c) Oracle – SQL.

d) Correct. MongoDB stores JSON-like documents.

e) SQLite – SQL-based.
---

28. Which data format is commonly used for data interchange?

Answer: b) JSON

a) DOCX – Document format.

b) Correct. JSON is lightweight, widely used in APIs.

c) EXE – Executable file.

d) MP3 – Audio.

e) PNG – Image.

---

29. A confusion matrix is used in:

Answer: c) Classification

a) Clustering – Different metrics.

b) Regression – Use RMSE, R².

c) Correct. It shows true/false positives/negatives.

d) Time series – Use forecasting metrics.

e) Dimensionality reduction – Not applicable.

---

30. Which is a common performance metric for regression?

Answer: d) RMSE

a) Accuracy – For classification.

b) Precision – Classification.

c) Recall – Classification.

d) Correct. RMSE = root mean squared error.

e) F1 Score – Classification.

---

31. A benefit of using cloud platforms for data science is:

Answer: b) Scalability

a) Hardware replacement – Not a direct benefit.

b) Correct. Cloud offers flexible compute/storage.

c) Offline data access – Often limited.

d) Guaranteed model accuracy – Depends on model, not platform.

e) Source code encryption – Not a main feature.

---

32. Which is an example of categorical data?

Answer: c) Eye color

a) Height – Numerical.

b) Age – Numerical.

c) Correct. Categories like blue, green.

d) Temperature – Continuous.

e) Weight – Continuous.
---

33. In k-means clustering, ‘k’ refers to:

Answer: b) Number of clusters

a) Features – Columns.

b) Correct. It defines how many groups to form.

c) Data points – Rows.

d) Iterations – Not related directly.

e) Distance metric – Used internally.

---

34. Which library is used for data visualization in Python?

Answer: b) Matplotlib

a) NumPy – Numerical ops.

b) Correct. Basic plotting library.

c) Scikit-learn – Machine learning.

d) TensorFlow – Deep learning.

e) Flask – Web apps.

---

35. Which of the following is an ensemble method?

Answer: c) Random Forest

a) Naive Bayes – Single model.

b) KNN – Lazy learner.

c) Correct. Random Forest uses many trees.

d) Logistic regression – Individual model.

e) PCA – Dimensionality reduction.

---

36. Which is NOT a type of machine learning?

Answer: e) Exhaustive learning

a) Supervised – Yes.

b) Unsupervised – Yes.

c) Reinforcement – Yes.

d) Semi-supervised – Yes.

e) Correct. Exhaustive learning is not a real type.

---

37. A data lake stores:

Answer: b) Raw, unprocessed data

a) Only structured data – False.

b) Correct. Data lakes store raw data in any format.

c) Only cleaned data – That’s a warehouse.

d) Tabular data only – False.

e) Secure documents – May include, but not purpose.

---

38. Which tool helps automate machine learning workflows?

Answer: b) AutoML

a) Excel – Manual.

b) Correct. AutoML automates model building/tuning.

c) MS Word – Not for ML.

d) Tableau – Visualization.

e) Hadoop – Processing platform.

---

39. Which library supports numerical computing in Python?

Answer: a) NumPy

a) Correct. Core numerical array ops.

b) Flask – Web.

c) Seaborn – Visualization.

d) Pandas – Dataframes.

e) PyTorch – Deep learning.

---

40. What is the role of the activation function in neural networks?

Answer: c) Introduces non-linearity

a) Reduces overfitting – Use regularization.

b) Initializes weights – Done separately.

c) Correct. Allows complex patterns.

d) Normalizes outputs – Use batchnorm.

e) Increases dataset size – Not related.

41. The bias-variance tradeoff refers to:

Answer: b) Model generalization

a) Ethical considerations – No.

b) Correct. Balance between underfitting (high bias) and overfitting (high variance).

c) Data storage costs – No relation.

d) Data collection methods – No.

e) Visualization clarity – No.

---

42. Which technique helps address class imbalance?

Answer: a) SMOTE

a) Correct. Synthetic Minority Over-sampling Technique generates synthetic samples.

b) Normalization – Scaling data, no class balance.

c) One-hot encoding – Encoding categorical vars.

d) Tokenization – NLP text splitting.

e) Aggregation – Summarizing data.

---

43. In NLP, stop word removal is an example of:

Answer: a) Data cleaning

a) Correct. Stop words are common words removed to reduce noise.

b) Model training – Not relevant here.

c) Visualization – No.

d) Deployment – No.

e) Hardware setup – No.

---

44. Which cloud service provides BigQuery analytics?

Answer: b) Google Cloud

a) AWS – Has Redshift.

b) Correct. BigQuery is Google Cloud’s data warehouse.

c) Azure – Has Synapse.

d) IBM Cloud – Different services.

e) Oracle – Different platform.

---

45. The primary goal of dimensionality reduction is to:

Answer: b) Simplify datasets

a) Increase data volume – No.

b) Correct. Reduce features to improve models and speed.

c) Create backups – No.

d) Improve storage – Side effect.

e) Encrypt data – No.

---

46. Which algorithm uses gradient boosting?

Answer: b) XGBoost

a) KNN – No boosting.

b) Correct. XGBoost is a gradient boosting framework.

c) K-means – Clustering.

d) SVM – Margin based.

e) Apriori – Association rules.

---

47. In data visualization, the “data-ink ratio” emphasizes:

Answer: b) Informative elements

a) Color variety – No.

b) Correct. Ratio of ink used for data vs. decoration.

c) 3D effects – Usually reduces data-ink ratio.

d) Animation – No.

e) Font styles – No.

---

48. Which phase converts raw data into analysis-ready format?

Answer: b) Preprocessing

a) Collection – Gathering raw data.

b) Correct. Cleaning, transforming data.

c) Modeling – Building models.

d) Deployment – Putting models into use.

e) Monitoring – Tracking model performance.

---

49. The term “MapReduce” is associated with:

Answer: b) Distributed processing

a) Data visualization – No.

b) Correct. It’s a programming model for distributed computing.

c) Neural networks – No.

d) Statistical testing – No.

e) Database design – No.

---
50. Which is NOT a characteristic of big data?

Answer: d) Validity

a) Volume – Yes.

b) Velocity – Yes.

c) Variety – Yes.

d) Correct. Validity is not a standard “V” of big data.

e) Veracity – Yes.

---

51. In computer vision, YOLO is used for:

Answer: b) Object detection

a) Image classification – No.

b) Correct. YOLO (You Only Look Once) detects objects in images.

c) Style transfer – No.

d) Data augmentation – No.

e) Noise reduction – No.

---

52. Which tool is used for containerization in deployment?

Answer: a) Docker

a) Correct. Docker packages applications in containers.

b) Tableau – Visualization.
c) Excel – Spreadsheet.

d) Hadoop – Big data processing.

e) Spark – Big data processing.

---

53. The F1-score combines which two metrics?

Answer: b) Precision & Recall

a) Accuracy & Precision – No.

b) Correct. Harmonic mean of precision and recall.

c) Recall & Specificity – No.

d) RMSE & MAE – Regression metrics.

e) Variance & Covariance – Statistical metrics.

---

54. Which data science role focuses on data pipelines?

Answer: b) Data Engineer

a) Business Analyst – Focuses on analysis, not pipelines.

b) Correct. Builds and maintains data workflows.

c) UX Designer – Interface design.

d) Network Admin – Infrastructure.

e) Security Specialist – Security.

---

55. The term “shuffling” in data processing refers to:

Answer: b) Randomizing data order

a) Data encryption – No.

b) Correct. Helps prevent order bias during training.

c) Deleting duplicates – Different.

d) Compression – No.

e) Visualization – No.

---

56. Which algorithm is unsupervised?

Answer: b) K-means

a) Linear Regression – Supervised.

b) Correct. K-means clusters unlabeled data.

c) SVM – Supervised.

d) Logistic Regression – Supervised.

e) Decision Trees – Supervised.

---

57. In time-series analysis, autocorrelation measures:

Answer: b) Relationship with past values

a) Data storage needs – No.

b) Correct. Correlation between time-lagged values.

c) Missing data – No.

d) Feature importance – No.

e) Model complexity – No.

---

58. Which Python library is used for deep learning?

Answer: c) TensorFlow

a) Pandas – Dataframes.

b) NumPy – Numerical.

c) Correct. TensorFlow is for deep learning.

d) Matplotlib – Visualization.

e) Scikit-learn – ML, but not deep learning focused.

---

59. The term “overfitting” occurs when a model:

Answer: a) Performs well on training data only

a) Correct. Model fails to generalize.

b) Generalizes to new data – Opposite.

c) Ignores important features – No.

d) Runs too slowly – No.

e) Requires less memory – No.

---

60. Which technique creates synthetic data for testing?

Answer: a) Bootstrapping

a) Correct. Resampling technique to create synthetic samples.

b) Imputation – Fills missing data.

c) Normalization – Scaling.

d) Tokenization – Text splitting.

e) Encryption – Data security.