Starting with Machine Learning on devices like MCU - Part 6 (Feature Generation)

Feature generation (or feature engineering) is a key step in building machine learning models because the right features can make or break model performance. There are several ways to generate features from data, depending on the type of data (numerical, categorical, text, images, time series, etc.) and the problem domain. Feature generation can be manual (domain-driven), statistical, interaction-based, or automated. The goal is always to create informative, discriminating, and compact representations of the data that help the model learn patterns efficiently. Some of the ways are,

• Mathematical / Statistical Transformations
• Categorical Feature Encoding
• Temporal / Date-Time Features
• Text Data Features
• Image / Audio Features
• Domain-Specific / Expert Features
• Feature Selection / Importance Driven
• Automated / Machine Learning Based Feature Generation