Introduction to Feature Engineering for ML: Transformation, Extraction, and Selection

GSB 109

In statistical machine learning, the data eventually has to boil down to appropriate numeric values, i.e., features, to be used by the Machine Learning (ML) models. The transformation process is called Feature Engineering, which consumes 70-80% of the ML workflow. In this session, we will discuss the basic techniques of feature transformation, feature extraction and feature selection. We will also cover various examples of engineering numeric, textual and categorical features along with important techniques for detecting and handling outliers and missing values.