Definition
The process of creating new input variables or transforming existing ones from the raw data to significantly improve the predictive performance of a machine learning model.
In short: The process of creating new input variables or transforming existing ones from the raw data to significantly improve the predictive performance of a machine learning model.