A literature review of feature selection methods

EasyChair Preprint 4949, version 1

Abstract

The process of accommodating data is limited by the evolution of hardware and technologies, and the current analytical tools are not sufficient enough to retrieve information from this current overwhelming flood of data. The agenda of feature selection is to choose a subset of features from the input space while reducing effects, from noise or irrelevant features, and still efficiently describe the input data that ends up in good prediction results. We have observed that the vast majority of papers we reviewed emphasizes on handling high-dimensional data with the help of human being interference. With having said that the authors, very often, came up with methods that are less computational than the methods that are currently available in the market. In this work, we present basic knowledge about feature selection methods, review a number of papers and discuss their disadvantages, and draw conclusions based on our review.  Finally, we suggest some future works which are worth to be worked on and investigate.

Keyphrases: Big Data, Data Mining, data preprocessing, feature selection methods, machine learning

@booklet{EasyChair:4949,
  author    = {Dyari M. Ameen M. Shareef and Ghader Ali Yosefi},
  title     = {A literature review of feature selection methods},
  howpublished = {EasyChair Preprint 4949},
  year      = {EasyChair, 2021}}