Machine Learning - Interview Questions

Beginner

1. Bias-variance tradeoff: how do under/overfitting manifest and how do you address them? (See Model Selection)
2. Data leakage: examples and prevention (pipelines, split before transform). (See Preprocessing)
3. Metrics: accuracy vs F1 vs ROC-AUC vs PR-AUC; when to use each? (See Evaluation)
4. Cross-validation: k-fold, stratified, group, time-series splits — when and why?

Intermediate

1. Feature scaling/encoding with ColumnTransformer and leakage-safe target encoding.
2. Class imbalance: thresholds, resampling, class weights, calibration.
3. Hyperparameter search: grid vs randomized vs Bayesian; early stopping.
4. Feature importance vs permutation importance; SHAP/LIME basics.

Advanced

1. Fairness: equalized odds vs demographic parity; tradeoffs. (See Ethics)
2. Productionization: online vs batch, blue/green/canary, shadow mode.
3. Monitoring: data/label drift, concept drift, retraining triggers. (See Monitoring)
4. Deep Learning: choosing CNN/RNN/Transformers for tasks; transfer learning.