Open AccessFeature Selection: An Assessment of Some Evolving Methodologies
Author(s) -
A. Abdul Rasheed
Publication year - 2021
Publication title -
türk bilgisayar ve matematik eğitimi dergisi
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.218
H-Index - 3
ISSN - 1309-4653
DOI - 10.17762/turcomat.v12i2.1802
Subject(s) - feature selection , computer science , feature (linguistics) , artificial intelligence , particle swarm optimization , simulated annealing , selection (genetic algorithm) , machine learning , variety (cybernetics) , data mining , philosophy , linguistics
Feature selection has predominant importance in various kinds of applications. However, it is still considered as a cumbersome process to identify the vital features among the available set for the problem taken for study. The researchers proposed wide variety of techniques over the period of time which concentrate on its own. Some of the existing familiar methods include Particle Swarm Optimisation (PSO), Genetic Algorithm (GA) and Simulated Annealing (SA). While some of the methods are existing, the emerging methods provide promising results compared with them. This article analyses such methods like LASSO, Boruta, Recursive Feature Elimination (RFE), Regularised Random Forest (RRF) and DALEX. The dataset of variant sizes is considered to assess the importance of feature selection out of the available features. The results are also discussed from the obtained features and the selected features with respect to the method chosen for study.