Feature selection is arguably the most challenging part of modeling that requires domain knowledge and good insights into the problem at hand. Nevertheless, properties of well-behaved features are as follows:
- Reusability: Features should be available for reuse in different models, applications, and teams.
- Transformability: You should be able to transform a feature with an operation, for example, log(), max(), or combine multiple features together with a custom calculation.
- Reliability: Features should be easy to monitor and appropriate unit tests should exist to minimize bugs or issues.
- Interpretability: To perform any of the previous actions, you need to be able to understand the meaning of features and interpret their ...