Feature engineering is the addition or construction of supplementary features to a dataset to improve model performance. Some of the approaches are outlined below.
Enrichment: supplementing a dataset with additional data from external sources, e.g., adding customer details to a dataset containing phone usage
Aggregations: calculating counts, sum, difference, average, most recent, minimum, maximum, range of a feature
Extracting features: e.g., extracting day of week (Monday, Tuesday, etc), time of day (morning, afternoon, etc) from datetime values, or extracting gender from name
Feature interaction: combining two or more features to produce a new effect or behavior of the original features. Aristotle’s predicate “The whole is greater than the sum of its parts,” applies here. For example, calculating body-mass index from weight and height, or computing polynomial combinations between features to enable linear methods to detect non-linear relationship between the features and the target.
