Feature Selection and Feature Extraction: Highlights

Abstract

In recent years, big data deluges have resulted in exciting data science opportunities. In particular, there is always a desire to extract the most from different data sources. To address it, a promising and recurring task is to perform feature selection and feature extraction. Specifically, the objective is to obtain the non-redundant and informative set of input features (also known as attributes or predictor variables) for downstream data science tasks. In this study, we highlight the existing approaches in both feature selection and feature extraction. In particular, benchmark comparisons are conducted for independent evaluations.