Why do data scientists perform feature engineering?
Show answer and explanation
Answer: B
The representation of the input strongly influences what a model can learn. Feature engineering creates informative signals, for example converting a timestamp into day-of-week and hour, deriving a ratio from two columns, or encoding categorical values appropriately. Better inputs can raise accuracy, reduce training time and make a simpler model viable. The work requires domain knowledge and must be applied consistently at training and scoring time to avoid skew.
Why the other options are wrong
- A. Releasing a model is deployment, a separate lifecycle stage.
- C. Monitoring live traffic is an operational task performed after deployment.
- D. Manual labelling produces targets for supervised learning; feature engineering shapes the inputs, not the labels.
Exam tip. Better inputs make patterns easier to learn.