Surrogate Sample-Assisted Particle Swarm Optimization for Feature Selection on High-Dimensional Data DOI

Xianfang Song,

Zhang Yon, Dunwei Gong

et al.

IEEE Transactions on Evolutionary Computation, Journal Year: 2022, Volume and Issue: 27(3), P. 595 - 609

Published: May 16, 2022

With the increase of number features and sample size, existing feature selection (FS) methods based on evolutionary optimization still face challenges such as "curse dimensionality" high computational cost. In view this, dividing or clustering spaces at same time, this article proposes a hybrid FS algorithm using surrogate sample-assisted particle swarm (SS-PSO). First, nonrepetitive uniform sampling strategy is employed to divide whole set into several small-size subsets. Regarding each subset unit, next, collaborative mechanism proposed space, with purpose reducing both cost search space PSO. Following that, an ensemble surrogate-assisted integer PSO proposed. To ensure prediction accuracy when evaluating particles, construction management designed. Since replaced by small units, SS-PSO significantly reduces particles in Finally, applied some typical datasets, compared six algorithms, well its variant algorithms. The experimental results show that can obtain good subsets smallest most datasets. All verify highly competitive method for high-dimensional FS.

Language: Английский

Particle Swarm Optimization: A Comprehensive Survey DOI Creative Commons
Tareq M. Shami, Ayman A. El‐Saleh, Mohammed Alswaitti

et al.

IEEE Access, Journal Year: 2022, Volume and Issue: 10, P. 10031 - 10061

Published: Jan. 1, 2022

Particle swarm optimization (PSO) is one of the most well-regarded swarm-based algorithms in literature. Although original PSO has shown good performance, it still severely suffers from premature convergence. As a result, many researchers have been modifying resulting large number variants with either slightly or significantly better performance. Mainly, standard modified by four main strategies: modification controlling parameters, hybridizing other well-known meta-heuristic such as genetic algorithm (GA) and differential evolution (DE), cooperation multi-swarm techniques. This paper attempts to provide comprehensive review PSO, including basic concepts binary neighborhood topologies recent historical variants, remarkable engineering applications its drawbacks. Moreover, this reviews studies that utilize solve feature selection problems. Finally, eight potential research directions can help further enhance performance are provided.

Language: Английский

Citations

686
Review of swarm intelligence-based feature selection methods DOI
Mehrdad Rostami, Kamal Berahmand, Elahe Nasiri

et al.

Engineering Applications of Artificial Intelligence, Journal Year: 2021, Volume and Issue: 100, P. 104210 - 104210

Published: Feb. 26, 2021

Language: Английский

Citations

366
Multi-objective particle swarm optimization with adaptive strategies for feature selection DOI
Fei Han,

Wen-Tao Chen,

Qing-Hua Ling

et al.

Swarm and Evolutionary Computation, Journal Year: 2021, Volume and Issue: 62, P. 100847 - 100847

Published: Feb. 6, 2021

Language: Английский

Citations

138
Evolutionary Multitasking for Feature Selection in High-Dimensional Classification via Particle Swarm Optimization DOI
Ke Chen, Bing Xue, Mengjie Zhang

et al.

IEEE Transactions on Evolutionary Computation, Journal Year: 2021, Volume and Issue: 26(3), P. 446 - 460

Published: July 26, 2021

Feature selection (FS) is an important preprocessing technique for improving the quality of feature sets in many practical applications. Particle swarm optimization (PSO) has been widely used FS due to being efficient and easy implement. However, when dealing with high-dimensional data, most existing PSO-based approaches face problems falling into local optima high-computational cost. Evolutionary multitasking effective paradigm enhance global search capability accelerate convergence by knowledge transfer among related tasks. Inspired evolutionary multitasking, this article proposes a PSO approach FS. The converts task several low-dimensional tasks, then finds optimal subset between these Specifically, novel generation strategy based on importance features developed, which can generate highly tasks from dataset adaptively. In addition, new mechanism presented, effectively implement positive results demonstrate that proposed method evolve higher classification accuracy shorter time than other state-of-the-art methods classification.

Language: Английский

Citations

114
Advances in nature-inspired metaheuristic optimization for feature selection problem: A comprehensive survey DOI
Maha Nssibi, Ghaith Manita, Ouajdi Korbaa

et al.

Computer Science Review, Journal Year: 2023, Volume and Issue: 49, P. 100559 - 100559

Published: May 22, 2023

Language: Английский

Citations

79
Novel Improved Salp Swarm Algorithm: An Application for Feature Selection DOI Creative Commons
Miodrag Živković, Cătălin Stoean, Amit Chhabra

et al.

Sensors, Journal Year: 2022, Volume and Issue: 22(5), P. 1711 - 1711

Published: Feb. 22, 2022

We live in a period when smart devices gather large amount of data from variety sensors and it is often the case that decisions are taken based on them more or less autonomous manner. Still, many inputs do not prove to be essential decision-making process; hence, utmost importance find means eliminating noise concentrating most influential attributes. In this sense, we put forward method swarm intelligence paradigm for extracting important features several datasets. The thematic paper novel implementation an algorithm branch machine learning domain improving feature selection. combination with metaheuristic approaches has recently created new artificial called learnheuristics. This approach benefits both capability selection solutions impact accuracy performance, as well known characteristic algorithms efficiently comb through search space solutions. latter used wrapper improvements significant. paper, modified version salp proposed. solution verified by 21 datasets classification model K-nearest neighborhoods. Furthermore, performance compared best same test setup resulting better number proposed solution. Therefore, tackles demonstrates its success benchmark

Language: Английский

Citations

76
SFE: A Simple, Fast, and Efficient Feature Selection Algorithm for High-Dimensional Data DOI
Behrouz Ahadzadeh, Moloud Abdar, Fatemeh Safara

et al.

IEEE Transactions on Evolutionary Computation, Journal Year: 2023, Volume and Issue: 27(6), P. 1896 - 1911

Published: Jan. 23, 2023

In this article, a new feature selection (FS) algorithm, called simple, fast, and efficient (SFE), is proposed for high-dimensional datasets. The SFE algorithm performs its search process using agent two operators: 1) nonselection 2) selection. It comprises phases: exploration exploitation. the phase, operator global in entire problem space irrelevant, redundant, trivial, noisy features changes status of from selected mode to nonselected mode. exploitation searches with high impact on classification results successful FS However, after reducing dimensionality dataset, performance cannot be increased significantly. these situations, an evolutionary computational method could used find more subset reduced space. To overcome issue, article proposes hybrid SFE-PSO (particle swarm optimization) optimal subset. efficiency effectiveness are compared 40 Their performances were six recently algorithms. obtained indicate that algorithms significantly outperform other can as effective selecting

Language: Английский

Citations

67
A Survey on Evolutionary Multiobjective Feature Selection in Classification: Approaches, Applications, and Challenges DOI
Ruwang Jiao, Bach Hoai Nguyen, Bing Xue

et al.

IEEE Transactions on Evolutionary Computation, Journal Year: 2023, Volume and Issue: 28(4), P. 1156 - 1176

Published: July 5, 2023

Maximizing the classification accuracy and minimizing number of selected features are two primary objectives in feature selection, which is inherently a multiobjective task. Multiobjective selection enables us to gain various insights from complex data addition dimensionality reduction improved accuracy, has attracted increasing attention researchers practitioners. Over past decades, significant advancements have been achieved both methodologies applications, but not well summarized discussed. To fill this gap, paper presents broad survey on existing research classification, focusing up-to-date approaches, current challenges, future directions. be specific, we categorize basis different criteria, provide detailed descriptions representative methods each category. Additionally, summarize list successful real-world applications domains, exemplify their practical value demonstrate abilities providing set trade-off subsets meet requirements decision makers. We also discuss key challenges shed lights emerging directions for developments selection.

Language: Английский

Citations

63
A feature selection approach based on NSGA-II with ReliefF DOI
Yu Xue,

Haokai Zhu,

Ferrante Neri

et al.

Applied Soft Computing, Journal Year: 2023, Volume and Issue: 134, P. 109987 - 109987

Published: Jan. 6, 2023

Language: Английский

Citations

57
Harnessing eXplainable artificial intelligence for feature selection in time series energy forecasting: A comparative analysis of Grad-CAM and SHAP DOI Creative Commons
Corne van Zyl, Xianming Ye, Raj Naidoo

et al.

Applied Energy, Journal Year: 2023, Volume and Issue: 353, P. 122079 - 122079

Published: Oct. 17, 2023

This study investigates the efficacy of Explainable Artificial Intelligence (XAI) methods, specifically Gradient-weighted Class Activation Mapping (Grad-CAM) and Shapley Additive Explanations (SHAP), in feature selection process for national demand forecasting. Utilising a multi-headed Convolutional Neural Network (CNN), both XAI methods exhibit capabilities enhancing forecasting accuracy model efficiency by identifying eliminating irrelevant features. Comparative analysis revealed Grad-CAM's exceptional computational high-dimensional applications SHAP's superior ability revealing features that degrade forecast accuracy. However, limitations are found with Grad-CAM including decrease stability, SHAP inaccurately ranking significant Future research should focus on refining these to overcome further probe into other methods' applicability within time-series domain. underscores potential improving load forecasting, which can contribute significantly development more interpretative, accurate efficient models.

Language: Английский

Citations

51