Feature Selection Techniques for Data Analysis and Decision Making in Interdisciplinary Areas: A Systematic Review

Abstract

The widespread accessibility of data collection tools and social media practices has facilitated an exponential proliferation of data in several multidisciplinary fields. Data from numerous sources can help industries make informed judgments and advance in their fields. Due to the unprecedented amount of data, effective decision-making necessitates excellent feature selection analysis for selecting high quality data. Before analyzing the data with distinct Machine Learning models choosing the relevant quality features is vital for converting data into actionable insight. Initially, this study aimed to shed light on feature selection(FS), which is essential for capturing high-quality data and making effective decisions due to its inclusive nature and universal applicability. So, considering FS as crucial, we conducted an exhaustive, detailed, and systematic review by collecting articles from different sectors. In addition, the work also discussed the FS state of the art, various strategies in terms of filter, wrapper, hybrid, embedded, voting, ensemble, ranking, and the clear-cut view of the FS evolution. Further, the study addressed challenges and limitations pertaining to FS and conducted a comparative study analysis. We also suggested thumb rules to follow in order to design a solid FS strategy that could help us enhance overall system behavior and efficacy equally across multiple data sets.

Keywords

• Coefficient determination
• feature engineering
• feature selection in multidisciplinary fields
• machine learning
• subset evaluation