Scribd
Upload
4 views2 pages

Enhancing Cuckoo Search Algorithm for Feature Selection

The study by Feizi-Derakhsh and Kadhim (2023) explores the optimization of feature selection using the Cuckoo Search Algorithm (CSA) enhanced by a sinusoidal chaotic map and Oppositional-Based Learning (OBL) for improved population initialization. This hybrid approach aims to address limitations in the conventional CSA, such as poor population diversity and premature convergence, by promoting better exploration-exploitation balance. The study's objectives include integrating the enhanced CSA with a Support Vector Machine classifier and evaluating its performance on real-world datasets against previous research.

Uploaded by

Sufyan
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
4 views2 pages

Enhancing Cuckoo Search Algorithm for Feature Selection

The study by Feizi-Derakhsh and Kadhim (2023) explores the optimization of feature selection using the Cuckoo Search Algorithm (CSA) enhanced by a sinusoidal chaotic map and Oppositional-Based Learning (OBL) for improved population initialization. This hybrid approach aims to address limitations in the conventional CSA, such as poor population diversity and premature convergence, by promoting better exploration-exploitation balance. The study's objectives include integrating the enhanced CSA with a Support Vector Machine classifier and evaluating its performance on real-world datasets against previous research.

Uploaded by

Sufyan
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd

Problem Statement

In a recent study by Feizi-Derakhsh and Kadhim (2023), the Cuckoo Search Algorithm (CSA)
was employed to optimize feature selection in various real-world datasets. The methodology
utilized the Chi-square test for preliminarily identifying optimal features based on their statistical
relevance to the target class and a sinusoidal chaotic map for adaptively adjusting the step size
(α) during the CSA's search process, particularly in local regions. This approach refined the
algorithm's search for optimal solutions by balancing exploration and exploitation.

While their method effectively reduced feature dimensionality and enhanced the CSA's search
efficiency, thereby improving overall performance, it had limitations. Notably, the conventional
random population initialization in the CSA hindered the algorithm's ability to thoroughly
explore the search space, often resulting in poor population diversity, premature convergence,
and susceptibility to stagnation, ultimately leading to suboptimal solutions.

To address these challenges, this study proposes integrating a hybrid of the sinusoidal chaotic
map with Oppositional-Based Learning (OBL) for enhanced population initialization. This
combination is expected to improve search diversity and convergence by leveraging the
sinusoidal chaotic map's ability to generate diverse and well-distributed initial solutions, along
with OBL's capacity to accelerate convergence by exploring opposite regions of the search space.
These enhancements aim to promote a more balanced exploration-exploitation process, improve
classification accuracy, and significantly reduce feature dimensions in real-world applications.
Aim and Objectives

The study aims to enhance the Cuckoo Search Algorithm's performance in feature selection by
integrating a hybrid of a sinusoidal chaotic map and Oppositional-Based Learning (OBL) for
improved population initialization.

The aim will be achieved through the following objectives:

i. Propose a hybrid approach combining the sinusoidal chaotic map with Oppositional-

Based Learning (OBL) in the CSA to enhance population initialization, improving

search diversity and convergence.

ii. Integrate the enhanced CSA with a Support Vector Machine (SVM) classifier for

feature selection and classification tasks.

iii. Evaluate the proposed model's performance using eight real-world benchmark datasets,

comparing it with the study by Feizi-Derakhsh and Kadhim (2023) and other relevant

research, focusing on classification accuracy and the number of selected features.

You might also like