Machine Learning interview questions and quick answers
Supervised vs unsupervised learning, overfitting, metrics and the classic models, explained simply.
Review 899 questions and short answers on Machine Learning. Start with real examples below, then search, practise and follow links to full explanations.
Machine Learning: questions worth revising first
These answers come from AI Tapri's existing question bank and published interview guides. Open each answer and read the linked in-depth guide for examples and context.
01A fraud model shows 99% accuracy. Why might it still be useless?
If only 1% of transactions are fraud, a model that always says 'not fraud' scores 99% accuracy while catching zero fraud. On imbalanced data, look at recall, precision, F1 or PR-AUC instead.
Read the full explanation02What are the main assumptions of linear regression?
Linearity between features and target, independent errors, constant error variance (homoscedasticity), roughly normal errors (needed mainly for confidence intervals), and no severe multicollinearity among features.
Read the full explanation03What is bootstrap sampling and the out-of-bag score in a random forest?
Each tree trains on a bootstrap sample: n rows drawn with replacement, containing about 63% of the unique rows. The remaining roughly 37% are out-of-bag for that tree and give a free validation estimate.
Read the full explanation04Why can't a random forest or decision tree extrapolate beyond the training range?
Tree leaves hold averages of training targets, so a prediction can never exceed the largest or fall below the smallest value seen in training. For trends that continue upward, such as sales growth, linear models or feature engineering work better.
Read the full explanation05How do you choose k in KNN, and what happens when k is very small or very large?
Small k (like 1) follows noise and overfits; very large k smooths too much and underfits, at the extreme predicting the majority class for everything. Choose k by cross-validation, and use an odd k for two-class problems to avoid ties.
Read the full explanation06How do class weights, oversampling and undersampling differ for imbalanced data?
Class weights make mistakes on the rare class cost more during training. Oversampling duplicates or synthesises minority rows; undersampling removes majority rows. Weights change no data, oversampling can overfit, undersampling can throw away useful rows.
Read the full explanation07What is model selection?
Choose algorithm, hyperparameters. Train multiple, compare validation performance. Cross-validation guides choice.
08Explain model deployment.
Move trained model to production. API endpoint, containerize, monitor. Versioning, A/B testing.
Find your next answer.
Pick a topic, search a keyword, or open a question to reveal the answer.
Machine Learning quick answers
Free to practise. No account needed. Made for quick revision, not rote memorisation.
Keep preparing for Machine Learning
Quick answers help with recall; detailed guides explain the reasoning, examples and common mistakes behind each answer.