DSE Cheat Sheet 2026

The 30 highest-yield DSE facts, distilled from real exam questions. Print it, save it as a PDF, or study it here — free, no sign-up.

1. What is the main advantage of using quantile transformation over standard normalization for highly skewed data? → It maps values to a uniform or normal distribution, making it robust to outliers
2. Which business intelligence tool is most widely used for creating interactive dashboards without coding? → Tableau
3. What does a Precision-Recall curve's Area Under the Curve (AUC-PR) near 0.5 indicate for a balanced dataset? → The model performs at chance level
4. Which visualization is most effective for displaying trends of multiple categories over time on the same chart? → Multi-line time-series chart
5. What is the purpose of a train/test split in machine learning? → To evaluate model performance on unseen data
6. In the context of clustering high-dimensional data, what is the 'curse of dimensionality' problem? → Distance metrics become less meaningful as all pairwise distances converge
7. In hypothesis testing, which of the following describes a Type I error? → Rejecting a true null hypothesis.
8. What is the primary analytical purpose of a heatmap in data science? → Visualizing magnitude across a two-dimensional matrix using color intensity
9. Which statement about bootstrap confidence intervals is TRUE? → They approximate the sampling distribution using resampling from observed data
10. In coreference resolution, what is the task the model must solve? → Clustering mentions in text that refer to the same real-world entity
11. When using Fuzzy C-Means clustering, what does the fuzziness parameter 'm' control? → The degree of overlap between cluster memberships
12. Which metric is most important when evaluating the performance of a big data storage system under write-heavy workloads? → Write throughput measured in MB/s or records/second
13. The IEEE Ethically Aligned Design framework prioritizes which of the following as its first principle? → Human well-being as the primary goal of AI development
14. The Kolmogorov-Smirnov test is primarily used to: → Test whether a sample comes from a specified distribution
15. Which metric measures the proportion of actual positives correctly identified by a classifier? → Recall
16. Which of the following is an example of a hyperparameter (not a learned parameter) in a neural network? → The learning rate used by the optimizer
17. In a Random Forest classifier, how does increasing the number of trees typically affect variance? → Decreases variance
18. Which technique helps identify the most important variables early in EDA without building a full model? → Computing correlation with the target variable or using mutual information scores
19. In gradient boosting for classification, what does each successive tree learn? → The residual errors (pseudo-residuals) of the previous ensemble
20. In a neural network, what is the purpose of the backpropagation algorithm? → Computing gradients of the loss with respect to each weight so they can be updated
21. Why is walk-forward (time series cross-validation) preferred over standard k-fold cross-validation for time series data? → It respects temporal ordering and prevents future data from leaking into the training set
22. What is the primary advantage of using gradient boosting over random forests? → It sequentially corrects errors from previous trees, often achieving lower bias
23. What is a confusion matrix used for? → Evaluating classification model performance
24. Shapley values in model explanation primarily help with evaluation by: → Attributing each feature's contribution to individual predictions fairly
25. Which architectural design is also referred to as a systolic array? → MISD
26. Which NLP task involves identifying the grammatical role of each word in a sentence, such as noun or verb? → Part-of-speech tagging
27. What does ETL stand for in data engineering? → Extract, Transform, Load
28. Which metric is most appropriate for evaluating a classifier on a heavily imbalanced dataset? → F1-Score or AUC-ROC
29. In a reliability diagram (calibration plot), a perfectly calibrated model's curve would: → Lie along the diagonal from (0,0) to (1,1)
30. In Named Entity Recognition (NER), what does the BIO tagging scheme stand for? → Beginning, Inside, Outside