Empowering African vultures optimizer using Archimedes optimization algorithm for maximum efficiency for global optimization and feature selection

Feature selection (FS) plays a pivotal role in data mining, presenting an optimization challenge that seeks to reduce feature size and enhance model generalization simultaneously. The expansive search space involved in this task often leads conventional optimization methods to produce suboptimal out...

Full description

Saved in:
Bibliographic Details
Published in:Evolving systems 2024-10, Vol.15 (5), p.1701-1731
Main Authors: Mostafa, Reham R., Hashim, Fatma A., El-Attar, Noha E., Khedr, Ahmed M.
Format: Article
Language:English
Subjects:
Accuracy
Algorithms
Artificial Intelligence
Complex Systems
Complexity
Convergence
Data mining
Datasets
Design optimization
Efficiency
Engineering
Feature selection
Global optimization
Heuristic
Machine learning
Optimization algorithms
Optimization techniques
Original Paper
Search process
User generated content
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Feature selection (FS) plays a pivotal role in data mining, presenting an optimization challenge that seeks to reduce feature size and enhance model generalization simultaneously. The expansive search space involved in this task often leads conventional optimization methods to produce suboptimal outcomes, impeding the pursuit of the most optimal global solution. This study presents a novel hybrid optimization approach, AVOA-AOA, which integrates the Archimedes Optimization Algorithm (AOA) with the African Vultures Optimization Algorithm (AVOA) specifically for numerical optimization and FS purposes. AVOA algorithm, inspired by the foraging behavior of vultures in Africa, is known for its simple yet effective design. However, it has some drawbacks, such as limited exploration capacity and early convergence due to minimal search process exploration. These drawbacks lead to an unbalanced search with an inability to bypass local solutions. To address these issues, the proposed AVOA-AOA method integrates AOA into the AVOA framework, enhancing the exploration phase and facilitating a dynamic transition between exploration and exploitation. The in-depth investigation and analysis of the performance of the proposed AVOA-AOA method are carried out using (1) twenty-nine CEC2017 benchmark functions and (2) nineteen datasets for FS problems. The experimental findings show that AVOA’s search tactics and convergence behavior have been vastly enhanced. The proposed AVOA-AOA achieves an average accuracy rate of 97.64% using the K-Nearest Neighbors (KNN) classifier and surpasses state-of-the-art results for 15 out of 19 datasets.
ISSN:1868-6478
1868-6486
DOI:10.1007/s12530-024-09585-6