Feature engineering transforms raw data into consistent, useful, and informative inputs for machine-learning algorithms. It belongs in the core ML pipeline, before modelling, and supports both technical data quality and the encoding of human insight. section 7.2
Turn the work into an executable delivery ticket:
- Create and review a feature-engineering plan with the team. section 7.2
- Implement a feature-engineering pipeline that transforms raw data for algorithm consumption. section 7.2
- Inspect repeated information, distribution skews, scale differences, and hierarchical data before modelling. section 7.2
- Design and add data augmentation where the available data or features are insufficient. section 7.2
Representing cyclic time
A model may not infer that January 1 follows December 31 when dates are represented as 1 and 12. Feature engineering can encode the calendar’s cyclical structure rather than leaving the algorithm with disconnected scalar values. section 7.2
Figure 7.1 Temperatures plotted by day of the year (data for Suffolk Wattlesham weather station, 2019, from UK Government).
Compare temperatures plotted by day of year with an alternative representation based on distance from midsummer. section 7.2
