Train with imbalanced sampled data
Company: LinkedIn
Role: Data Scientist
Category: Machine Learning
Difficulty: medium
Interview Round: Technical Screen
You are training a binary classifier on a very large dataset where the positive class is rare. Because the full dataset is too large to train on directly, you plan to draw a sample and train a tree-based model.
Explain how you would:
1. Handle class imbalance during training.
2. Verify that the sampled training data is representative of the full population.
3. Validate that a model trained on the sample generalizes to the full dataset.
4. Prevent overfitting in a tree-based model.
5. Choose evaluation metrics, especially when the classes are highly imbalanced.
Overview: This question evaluates competency in handling class imbalance, designing representative sampling strategies, verifying sample-to-population generalization, preventing overfitting in tree-based models, and selecting evaluation metrics for highly imbalanced binary classification.
Read the full LinkedIn Data Scientist interview experience this question came from