MS-DS Master of Data science Cheat Sheet 2026

The 30 highest-yield MS-DS Master of Data science facts, distilled from real exam questions. Print it, save it as a PDF, or study it here — free, no sign-up.

1. In the MapReduce programming model, what does the 'Shuffle and Sort' phase accomplish? → It groups all values by key and delivers them sorted to each reducer
2. What is the primary risk of using a dual-axis chart? → It can mislead viewers by implying a correlation between unrelated variables
3. In Gaussian Mixture Models (GMM), what algorithm is typically used to estimate the model parameters? → Expectation-Maximization (EM)
4. In a Gaussian Mixture Model (GMM), what algorithm is typically used to estimate the model parameters? → Expectation-Maximization (EM)
5. A researcher notices that their linear regression model has residuals that fan out as predicted values increase. What assumption is being violated? → Homoscedasticity
6. What is the main advantage of AdaBoost over a single decision tree? → It sequentially focuses on misclassified samples to improve overall accuracy
7. In pandas, what is the effect of setting the parameter how='outer' in pd.merge()? → It returns all rows from both DataFrames, filling NaN where there is no match
8. When making predictions, trees examine each set of data's . → homogeneity
9. Which visualization is most appropriate for examining the relationship between two continuous variables in exploratory data analysis? → Scatter plot
10. Which technique does Apache Spark use to optimize query execution plans in Spark SQL? → Catalyst optimizer
11. Which technique helps address class imbalance in a binary classification dataset? → SMOTE (Synthetic Minority Over-sampling Technique)
12. What is early stopping in neural network training? → Halting training when validation loss stops improving to prevent overfitting
13. Which regularization technique specific to transformer training helps stabilize optimization by normalizing activations before each sublayer's computation? → Pre-layer normalization (Pre-LN) applied before each sublayer
14. When applying a pipeline in scikit-learn, why should you fit the preprocessor only on training data and then transform both training and test data? → To prevent data leakage from the test set into the model
15. What is the primary purpose of cross-validation in machine learning model development? → To estimate how well the model generalizes to unseen data
16. What is the key assumption of the Naive Bayes classifier? → All features are conditionally independent given the class label
17. In the Transformer decoder, why is masked self-attention used during training? → To prevent the model from attending to future tokens, preserving autoregressive generation
18. In survival analysis, censored observations occur when: → A subject leaves the study or the study ends before the event is observed
19. Identify the error in the statement. → Adding squared terms makes it twice continuously differentiable at the knot points
20. In feature engineering, what is the purpose of creating interaction terms between two variables? → To capture the combined effect of two features that may not be represented by either alone
21. A random sample of size n=36 is drawn from a population with σ=12. What is the standard error of the mean? → 2
22. Which property ensures that a maximum likelihood estimator (MLE) achieves the Cramér-Rao lower bound asymptotically? → Asymptotic efficiency
23. When performing multiple hypothesis tests simultaneously, the Bonferroni correction adjusts α by: → Dividing α by the number of tests
24. When performing EDA on a dataset with 50 features, which technique best identifies groups of highly correlated predictors? → Clustered heatmap of the correlation matrix
25. Which cross-validation strategy is most appropriate when the dataset has a temporal ordering? → Time-series split (walk-forward validation)
26. In a convolutional neural network (CNN), what is the primary purpose of a pooling layer? → Reduce spatial dimensions and computation
27. Which of the following statements about random forest is accurate? → Random forest are difficult to interpret but often very accurate
28. Which Apache HBase concept ensures that writes are not lost if a RegionServer crashes before flushing to disk? → Write-Ahead Log (WAL)
29. Which hyperparameter in DBSCAN determines the minimum number of points required to form a dense region (core point)? → minPts
30. A p-value of 0.001 in a study with n = 10,000 primarily reflects: → A very small effect size that is detectable only due to the huge sample