What is the cost function of Logistic Regression
Data Science Consultant Interview Questions
40,327 data science consultant interview questions shared by candidates
How would you measure success of some future campaign? If it went wrong in a particular way, what would you look at? What would you recommend to remedy?
The questions were pretty much based on my CV and my experience.
They asked me to go in detail on certain research projects and work experiences, and asked me why I chose that particular methodology, what difficulties I found, what would I have done if I had a different data, etc.
We have 6 columns of data. How to compare the first 3 to the last 3 ? I had to use the coding note pad to write the codes.
Basic python manipulations, O(n) notation.
1. Started with a detailed explanation of a past project - what was the business question, how did you come up with the solution, what was your hypothesis, how did you design the A/B test, why did you make certain choices, what was the result etc. Prepare 1-2 examples from your past, where you can talk in depth about the technical elements of your project. 2. Let's say we have a dataset with attributes for a house (Sq footage, locality etc) and house price. How will you predict the house price from these attributes? (Build a multiple regression model) 3. For this multiple regression model, explain the end-to-end process. What steps will you take before building the model, how will you impute missing values, how will you handle outliers etc. What are the underlying assumptions of a regression model? 4. Once the model is built, how will you infer the relationship (sign and magnitude) between the house attributes and house price. How will you explain it to someone that's not a technical person? 5. For the regression coefficients, how will you interpret them, (p-values, confidence interval etc). How will you explain a p-value to a layman 6. Next question was about "how will you segment customers" in order to serve a business requirement, such as determining which customers to show a given ad (I answered with clustering, because the business problem wasn't very specific, he just described it very generally) 7. For clustering, how does it work, how to choose the value of K in k-means. I also said we can use Gaussian mixture models for clustering, which he didn't seem to know because he asked me to clarify what I mentioned. There might have been a few more questions that I don't remember, but the theme of the interview was to check how well you know the basics of Stats/ML. I believe I answered most of the questions correctly so to receive the feedback that I wasn't up to the mark technically seemed like a case of Google not wanting to reveal the real reason, whatever it was. Either way, make sure you confirm the format of the interview with the recruiter. Because I was already interviewing with other companies, I had brushed up on my Stats/ML basics, but you might not be as lucky. Good luck!
Tell me about YouTube abuse. that's all I remember, Glassdoor forced me to contribute
Questions came from three categories. 1) Coding of simple data-structure realted questions. 2) SQL questions 3) Statistics/ probability theory questions (these where somewhat to verify basic intuition of these fields rather seeking written formal answers)
Write an equation to optimize the marketing spend between Facebook and Twitter campaigns.
Viewing 1081 - 1090 interview questions