Glossary term
Glossary term
Foundations
A feature not present among the input features, but assembled from one or more of them. Methods for creating synthetic features include the following:
Bucketing a continuous feature into range bins.
Creating a feature cross.
Multiplying (or dividing) one feature value by other feature value(s) or by itself. For example, if a and b are input features, then the following are examples of synthetic features:
ab
a2
Applying a transcendental function to a feature value. For example, if c is an input feature, then the following are examples of synthetic features:
sin(c)
ln(c)
Features created by normalizing or scaling alone are not considered synthetic features.
T
Created for this library
A risk modeling team derives synthetic features like debt-to-income ratio from raw fields because ratios capture risk more directly than the raw values.
A retail forecasting team derives synthetic features like rolling average sales and price-to-competitor ratios to improve baseline models.
A churn team derives synthetic features like average sessions per week and recent-tenure flags to capture customer behavior shifts.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License