AIExplainer
Machine Learning Intermediate 2 min read

What is synthetic feature?

A feature that is engineered or constructed from existing features to improve model performance or reduce dimensionality

In machine learning, a synthetic feature is a new feature that is created by combining or transforming existing features. This is done to help the model learn more complex relationships between the data and improve its performance. Synthetic features can be created using various techniques such as polynomial transformations, interaction terms, or dimensionality reduction methods.

Think of synthetic features like cooking a new recipe. You take existing ingredients (features) and combine them in a new way to create something new and more flavorful (synthetic feature) that enhances the overall dish (model performance).

For example, in a model that predicts house prices, a synthetic feature could be created by multiplying the number of bedrooms by the square footage of the house. This new feature can help the model capture the relationship between the size and layout of the house and its price.

Synthetic features are used in machine learning to improve the accuracy and robustness of models. They can help to reduce overfitting, improve generalization, and increase the interpretability of the model.

One common misconception is that synthetic features are always better than the original features. However, this is not always the case, and the creation of synthetic features should be guided by domain knowledge and careful evaluation of the model's performance.

The concept of synthetic features has been around for decades and has its roots in feature engineering and dimensionality reduction techniques. With the advent of machine learning and deep learning, the use of synthetic features has become more widespread and automated.

engineered feature constructed feature derived feature

Three products for different needs — explore what’s relevant to you.