Deep Learning Model-Based on Feature Selection Algorithm

Research areas:

• Volume 10,Issue 3,Nov. 2021

Year:

2021

Type of Publication:

Article

Keywords:

Features Selection, Wrapper Approach, Filter Approach, Deep Learning, Particle Swarm Optimization

Authors:

Alaa Eisa; Nora EL-Rashidy; Hazem M. El-bakry; Samir Abdelrazek

Journal:

IJAIM

Volume:

10

Number:

3

Pages:

23-32

Month:

November

ISSN:

2320-5121

Abstract:

In these days, data grows rapidly with high dimensionality, so it is vital to provide some techniques to reduce this dimensionality. Feature selection is an important topic in the field of data mining. It is used for selecting important features to improve performance and reducing training time through dimensionality reduction. This paper provides a comprehensive overview of features selection algorithms. A comparison between various features selection approaches is introduced in this paper. The main contribution of this paper is to clarify the improvement on classification performance that features selection algorithms achieve. The authors provide a comparison between the proposed model DLPSO and the deep learning algorithm to show its performance. The performance metrics are accuracy rate, precision rate, classification error, relative error, and absolute error. The results show that the proposed model achieves high accuracy with 99.11 % with a minimum classification error of 0.89%.

Full text: IJAIM_646_FINAL.pdf