Indian Journal of Science and Technology

Abstract

Objectives: This study summarizes the feature selection process, its importance, different types of feature selection algorithms such as Filter, Wrapper and Hybrid. Moreover, it analyses some of the existing popular feature selection algorithms through a literature survey and also addresses the strengths and challenges of those algorithms. Methods/ Statistical Analysis: When there are many methods are in hand to be obtained, then Review of Literature is the best approach to learn about existing methods before going for a new model. Findings: Feature selection is a predominant preprocessing strategy in Data Mining, which helps in advancing the performance of mining, by selecting only the relevant features and avoiding the redundant features. There are plenty Feature Selection algorithms developed and used by most researchers. But still it is an emerging area in machine learning to be focused for data mining and analysis process for pattern recognition. Many feature selection algorithms confront severe challenges in terms of effectiveness and efficiency, because of recent increase in data variety and velocity. Different types of feature selection algorithms are available in literature such as Filter based, Wrapper based and Hybrid algorithms. Moreover, analyses some of the existing popular feature selection algorithms through a literature survey, also addresses the strengths and challenges of those algorithms. Application/Improvements: There is a need for an effective unified framework, which should provide feature selection for any size of dataset without noisy data, low computational complexity and highest accuracy. 

Keywords: Classification, Data Mining, Feature Selection, Filter, Hybrid, Wrapper