Some Data Science Interview Questions (by Me and Former Colleagues at SAS)

What is the curse of dimensionality and how should one deal with it when building machine-learning models?
Why is a comma a bad record separator/delimiter?
Explain the difference between a compiled computer language and an interpreted computer language.
How do you determine “k” for k-means clustering? Or, how do you determine the number of clusters in a data set?
What’s more important: predictive power or interpretability of a model?
Explain finite precision. Why is finite precision a problem in machine learning?
Explain the “bias-variance trade-off” and why it is fundamental to machine learning.
Describe a recent use of logistic regression.
Give examples of data cleaning techniques you have used in the past.
What subjects would you include in a one-day data science crash course? And why?
Describe a situation where you had to decide between two different types of analyses – and why you chose the one you did.
Explain the benefits of test-driven software development; or explain the benefits of unit testing.

Describe an analysis you have recently completed, including strategies and findings. How were the findings used by the business? (This can be from a student research project or thesis if the candidate is a recent graduate.)
Explain to the leaders of this company what model lift is and why they should care.
How do you identify and overcome obstacles (during projects, with customers, with decision makers, etc.)?
Tell me about a project you worked on that succeeded in part because of the way results were communicated.
What is your favorite data visualization book or blog? And why?
How would you design a chart or graph for a color-blind audience?
Tell me a compelling story about data that you have analyzed.
Explain to a business analyst the trade-off between the predictive power and the interpretability of a model – and why this matters.

jphall663 / ds_interview_qs