Systems Analyst: What is feature engineering and how do you approach it?

Feature engineering — transforming raw data into ML model inputs.

Techniques:

• Numerical: Scaling, normalization, binning.
• Categorical: One-hot encoding, target encoding, embedding.
• Temporal: Day of week, hour, seasonality.
• Text: TF-IDF, word embeddings, n-grams.
• Aggregations: Rolling averages, counts, ratios.

Process:

1. Understand business context and data.
2. Explore data (distributions, correlations).
3. Generate candidate features.
4. Select impactful features.
5. Validate with cross-validation.

Feature stores:

• Centralized feature repository.
• Consistent features across training and serving.
• Feature versioning and lineage.
• Online + offline feature serving.

Tools:

• Feast, Tecton, Hopsworks.
• Featuretools (automated).

Best practices:

• Avoid data leakage (future info in training).
• Document feature definitions.
• Monitor feature drift in production.