An input feature is a variable describing an example for a machine-learning model. Features turn observations such as customer behavior into information a clustering or classification model can use. section 4.1.5 section 4.2.2
Useful features are selected for relevance, interpretability, and practical actionability:
- Choose domain-relevant variables rather than allowing a model to prioritize every available pattern. section 4.1.5
- For customer clustering, useful features can include purchase frequency, average order value, browsing behavior, preferred product categories, and response to recommendations. section 4.1.5
- Treat scaling and repeated refinement as part of feature preparation, and question variables that add little meaningful insight. section 4.1.5
- In classification, features such as user profiles and behavior help predict labels, while feature contributions can support interpretation. section 4.2.2
A customer-segmentation input
A marketing model might represent a customer with purchase frequency, average order value, browsing behavior, preferred categories, and response to recommendations. The resulting clusters or predicted segment are only as useful as the variables chosen to represent that customer. section 4.1.5 section 4.2.2
