Question 1

Explain the bias-variance trade-off.

Accepted Answer

Bias is error from a model that's too simple (underfitting); variance is error from a model too sensitive to training data (overfitting). Reducing one often raises the other; aim for the lowest total error on unseen data.

Question 2

How do you detect and prevent overfitting?

Accepted Answer

Detect it when training performance is much better than validation. Prevent it with more data, simpler models, regularisation (L1/L2), cross-validation, early stopping and pruning for trees.

Question 3

Precision vs recall - when do you prioritise each?

Accepted Answer

Precision when false positives are costly (marking real emails as spam); recall when false negatives are costly (missing fraud or disease). F1 balances both.

Question 4

Your fraud dataset has 1% fraud. What do you do?

Accepted Answer

Don't use accuracy - 99% is easy. Use precision-recall metrics, class weights or resampling, pick a threshold by business cost, and check performance on a realistic validation set.

Question 5

What is data leakage?

Accepted Answer

When information unavailable at prediction time sneaks into training - for example a feature computed using the future, or scaling before the train-test split. Split first; build features only from the past.

Question 6

How would you design an A/B test for a new checkout button?

Accepted Answer

Define one primary metric (conversion), compute the sample size for the smallest effect you care about, randomise users, run for full weeks, avoid peeking early, then test significance and check guardrail metrics.

Question 7

Random forest vs gradient boosting?

Accepted Answer

Random forests average many independent deep trees - robust, less tuning. Gradient boosting builds trees in sequence, each fixing the last one's errors - often more accurate but needs careful tuning.

Question 8

What does a p-value mean?

Accepted Answer

The probability of seeing a result at least this extreme if there were truly no effect. It is not the probability that the hypothesis is true.

Question 9

How do you explain a model's predictions to business people?

Accepted Answer

Feature importance or SHAP values for which factors matter, simple examples of individual predictions, and the impact in business terms - money saved, customers retained.

Question 10

Walk me through a project end to end.

Accepted Answer

Problem, data, baseline, features, model, evaluation metric and the result compared with the baseline. Say what you'd improve with more time.

Question 11

What do data scientist interviews include?

Accepted Answer

Usually statistics and ML theory, a coding round in Python or SQL, a case study or take-home, and a discussion of past projects.

Question 12

How should a fresher prepare for data science interviews?

Accepted Answer

Learn statistics and core ML well, practise SQL and pandas, and complete two or three projects where you can explain the metric, the baseline and why your model is better.

Question 13

Is deep learning asked in data scientist interviews?

Accepted Answer

For general data scientist roles, mostly basics. Deep learning matters more for ML engineer, computer vision or NLP roles.

Data scientist interview questions and answers

What to revise before the interview

1.Explain the bias-variance trade-off.

2.How do you detect and prevent overfitting?

3.Precision vs recall - when do you prioritise each?

4.Your fraud dataset has 1% fraud. What do you do?

5.What is data leakage?

6.How would you design an A/B test for a new checkout button?

7.Random forest vs gradient boosting?

8.What does a p-value mean?

9.How do you explain a model's predictions to business people?

10.Walk me through a project end to end.

Frequently asked questions

Company interview guides for this role

Related roles and topics

Practise these questions for your next interview