Article
Computer Science, Artificial Intelligence
Lanlan Kang, Wenliang Cao, Ruey-Shun Chen, Yeh-Cheng Chen
Summary: This paper proposes a novel intelligent algorithm based on particle swarm optimization, called ARDDEA, to address prediction problems in a dynamic environment. The algorithm monitors environmental changes, avoids search overlapping of sub-populations with a new exclusion strategy, and enhances search capability through a new dual-drive dynamic updating equation.
APPLIED SOFT COMPUTING
(2021)
Article
Computer Science, Artificial Intelligence
Salihu A. Abdulkarim, Andries P. Engelbrecht
Summary: Several studies have applied particle swarm optimization algorithms to train neural networks for time series forecasting, with good performance results. This study introduces a dynamic PSO algorithm for training NNs in forecasting non-stationary time series, outperforming standard PSO and Rprop algorithms. These findings suggest the potential of dynamic PSO in real-world forecasting applications.
NEURAL COMPUTING & APPLICATIONS
(2021)
Article
Thermodynamics
Navid Kardani, Abidhan Bardhan, Pijush Samui, Majidreza Nazem, Panagiotis G. Asteris, Annan Zhou
Summary: This study introduces hybrid adaptive neuro swarm intelligence techniques for predicting thermal conductivity of unsaturated soils. The ANN-APSO model showed the best prediction performance, indicating its potential as an alternate solution for real-time engineering problems.
INTERNATIONAL JOURNAL OF THERMAL SCIENCES
(2022)
Article
Automation & Control Systems
Gilberto Rivera, Raul Porras, J. Patricia Sanchez-Solis, Rogelio Florencia, Vicente Garcia
Summary: This paper introduces a novel metaheuristic called Outranking-based Particle Swarm Optimization (O-PSO) for addressing the multi-objective Unrelated Parallel Machine Scheduling Problem. O-PSO is an optimization algorithm that combines particle swarm optimization with the preferences of the Decision Maker (DM) expressed in a fuzzy relational system based on ELECTRE III. Unlike other multi-objective metaheuristics, O-PSO focuses on finding the Region of Interest (RoI) instead of approximating a sample of the complete Pareto frontier. The efficiency of O-PSO is validated through experiments on synthetic instances and a real-world case study, showing its capability of generating high-quality solutions and supporting multicriteria decision analysis.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Cuie Yang, Yiu-ming Cheung, Jinliang Ding, Kay Chen Tan
Summary: This article proposes a hybrid ensemble approach to address the concept drift-tolerant transfer learning problem, adapting target domain models to new environments through class-wise weighted ensemble. The approach assigns weight vectors for classifiers from previous data chunks, allowing each class of current data to leverage historical knowledge independently.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Yi Liu, Gengsong Li, Xiang Li, Wei Qin, Qibin Zheng, Xiaoguang Ren
Summary: High dimensional imbalanced missing data classification is a challenging problem that traditional algorithms struggle with. To address this, a novel method called Hybrid Classification Approach based on Particle Swarm Optimization (HCPSO) is proposed. HCPSO integrates feature selection, resampling, and imputation, and uses particle swarm optimization to optimize these parameters. Evaluation using six algorithms, eleven datasets, and four performance indicators shows significant improvement in HCPSO's performance compared to other methods, with average improvements of 13.02%, 18.95%, 20.25%, and 28.63% for accuracy, F1, AUC, and Gmean, respectively. The experiments also demonstrate HCPSO's robustness.
ELECTRONICS LETTERS
(2023)
Article
Chemistry, Analytical
Hailun Xie, Li Zhang, Chee Peng Lim, Yonghong Yu, Han Liu
Summary: This research proposes two PSO variants for feature selection tasks to overcome shortcomings of the original PSO model. The first variant includes four operations, while the second enhances the first with four new strategies. Compared with other methods, these models demonstrate statistical superiority on 13 datasets.
Article
Chemistry, Analytical
Suganya Selvaraj, Eunmi Choi
Summary: This paper proposes an improved PSO algorithm, called dynamic sub-swarm PSO, for text document clustering problems. The experimental results show that this algorithm outperforms standard PSO and K-means algorithms in terms of purity and execution time.
Article
Materials Science, Multidisciplinary
Ryan Lye, Chris Bennett, James Rouse, Giuseppe Zumpano
Summary: Abradable coatings protect blades from damage during blade-casing interactions in gas turbine engines. This paper presents a method to rapidly estimate the properties of abradable coatings using simulated Rockwell hardness testing, and demonstrates the differences in contact forces and dominant frequencies during rub events through a series of simulations.
MATERIALS TODAY COMMUNICATIONS
(2022)
Article
Engineering, Industrial
Fehmi Burcin Ozsoydan, Ilker Golcuk
Summary: This paper introduces a cooperative approach using swarm intelligence algorithm and linear programming solver to solve the capacitated facility location problem. The proposed strategy decomposes the problem into two sub-problems and utilizes an improved algorithm and optimization model for solving. The experimental results show promising performance of the proposed strategy.
INTERNATIONAL JOURNAL OF PRODUCTION RESEARCH
(2023)
Article
Computer Science, Artificial Intelligence
Junhua Hu, Xiangzhu Ou, Pei Liang, Bo Li
Summary: This paper proposes a CART model based on particle swarm optimization to help patients choose between immunotherapy and cryotherapy for treating warts. The model is more accurate and concise than traditional methods, and can reduce medical costs and improve healing quality for patients and doctors. Experimental results show that the proposed model outperforms other classification methods and optimization algorithms.
COMPLEX & INTELLIGENT SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Ahmed Al Hilli, Mohanad Al-Ibadi, Ahmed M. Alfadhel, Sameer Hameed Abdulshaheed, Ahmed Hassan Hadi
Summary: This paper discusses the path planning problem in cases where 2D obstacles change their areas in stochastic form. The proposed approach utilizes Particle Swarm Optimization to find a path that is simultaneously safe and short, showing better effectiveness compared to traditional methods.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Ilker Golcuk, Fehmi Burcin Ozsoydan, Esra Duygu Durmaz
Summary: This paper introduces an improved Arithmetic Optimization Algorithm (AOA) for training artificial neural networks (ANNs) in dynamic environments. The proposed algorithm optimizes the connection weights and biases of the ANN under concept drift, outperforming state-of-the-art metaheuristic optimization algorithms in training ANNs for dynamic classification tasks. The findings demonstrate the potential of the improved AOA for dynamic data-driven applications.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Computer Science, Hardware & Architecture
Dhruba Jyoti Kalita, Vibhav Prakash Singh, Vinay Kumar
Summary: This paper proposes a novel knowledge-based approach using particle swarm optimization (PSO) to optimize the hyper-parameters of support vector machine. The approach includes a knowledge transfer module and a drift detection module to generate and transfer knowledge between consecutive time instances. Experimental results show that the proposed approach significantly reduces the average execution time compared to the traditional PSO method.
JOURNAL OF SUPERCOMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Jingwei Too, Ali Safaa Sadiq, Seyed Mohammad Mirjalili
Summary: This paper proposes a novel conditional opposition-based particle swarm optimization algorithm for feature selection. By introducing opposition-based learning and conditional strategy, the performance of the particle swarm optimization algorithm is improved. Experimental results demonstrate that the proposed approach not only achieves high prediction accuracy but also yields small feature sizes.
CONNECTION SCIENCE
(2022)
Article
Nuclear Science & Technology
Robert Nshimirimana, Ajith Abraham, Gawie Nothnagel, Andries Engelbrecht
Summary: This paper describes a simplified approach to optimizing the radiography process using a virtual environment, utilizing ray tracing technique and particle swarm optimization routine to calculate and optimize the design parameters of the radiography system. The method provides a straightforward virtual environment for basic radiography training and experimental planning.
NUCLEAR TECHNOLOGY
(2021)
Article
Computer Science, Artificial Intelligence
A. P. Engelbrecht, P. Bosman, K. M. Malan
Summary: This study explores the correlations between the characteristics of optimization problems and the behaviors of swarm-based algorithms, revealing links between specific problem features and algorithm behaviors. The research uses fitness landscapes characteristics and diversity rate-of-change to quantify the features of problems and algorithms.
Article
Computer Science, Information Systems
Mathys Ellis, Anna S. Bosman, Andries P. Engelbrecht
Summary: The article discusses the challenges faced by streamed data classifiers and proposes methods for analyzing the environment and difficulty of SDCP.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
Pawel Jocko, Beatrice M. Ombuki-Berman, Andries P. Engelbrecht
Summary: This study introduces archive management approaches for dynamic multi-objective optimisation problems using the multi-guide particle swarm optimisation (MGPSO) algorithm, which achieves efficient tracking of the changing Pareto-optimal front by proposing alternative archive update strategies.
SWARM INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Kyle Erwin, Andries Engelbrecht
Summary: Portfolio optimization is a multi-objective optimization problem that involves balancing risk and profit. Single-objective optimization methods, such as set-based particle swarm optimization (SBPSO), have been proposed to address this problem. This paper introduces the first multi-objective approach to SBPSO and compares its performance with other multi-objective algorithms. The results show that SBPSO is competitive with multiple runs, and the proposed multi-objective SBPSO achieves a more diverse set of optimal solutions.
Article
Computer Science, Artificial Intelligence
Kyle Erwin, Andries Engelbrecht
Summary: This paper provides a comprehensive review of over 140 papers that have applied evolutionary and swarm intelligence algorithms to the portfolio optimization problem. The papers are categorized based on the type of portfolio optimization problem considered and further classified into single-objective and multi-objective approaches. The various portfolio models used, as well as the constraints, objectives, and differences between them, are also discussed in detail. Based on the findings, guidance for future research in portfolio optimization is provided.
Article
Computer Science, Artificial Intelligence
Christiaan Scheepers, Andries Engelbrecht
Summary: This article investigates the shortcomings of Fonseca and Fleming's attainment surfaces and analyzes the quantitative measure based on attainment surfaces introduced by Knowles and Corne. The analysis reveals that the results of the Knowles and Corne approach are biased by the shape of the attainment surface. The article proposes improvements for bi-objective Pareto-optimal front (POF) comparisons and introduces a multi-objective optimization algorithm performance measure called the porcupine measure based on attainment surfaces. A computationally optimized version of the porcupine measure is presented and empirically analyzed.
Proceedings Paper
Computer Science, Artificial Intelligence
Thamsanqa Mlotshwa, Heinrich van Deventer, Anna Sergeevna Bosman
Summary: This paper examines the impact of loss function selection on the performance of artificial neural networks in supervised machine learning, comparing the Cauchy loss function to the mean squared error loss function on various regression problems.
ARTIFICIAL INTELLIGENCE RESEARCH, SACAIR 2022
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Isak Potgieter, Christopher W. Cleghorn, Anna S. Bosman
Summary: This study investigates the use of local optima network (LON) analysis to characterise and visualise the neural architecture space. The results indicate that LONs may provide a viable paradigm for analysing and optimising neural architectures.
2022 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN)
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Jarrod Goschen, Anna S. Bosman, Stefan Gruner
Summary: The ongoing progress in computational intelligence has increased the desire to apply CI techniques for improving software engineering processes, particularly software testing. Existing automated software testing techniques focus on using search algorithms to discover input values that achieve high execution path coverage. This paper introduces a novel genetic programming framework that evolves micro-programs to efficiently explore the input parameter domain of software components, rather than just focusing on input values.
2022 IEEE CONGRESS ON EVOLUTIONARY COMPUTATION (CEC)
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Stefan van Deventer, Andries Engelbrecht
Summary: A dynamic optimization method for training neural networks to predict trading behavior in the financial stock market is proposed and shown to significantly outperform static methods on a selection of South African stocks.
ADVANCES IN COMPUTATIONAL INTELLIGENCE (IWANN 2021), PT II
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
James Faure, Andries Engelbrecht
Summary: This paper proposes an approach to automate diagnosis of impacted teeth by analysing panoramic radiographs and training a convolutional neural network. Empirical results illustrate good performance in predicting impacted teeth.
ADVANCES IN COMPUTATIONAL INTELLIGENCE, IWANN 2021, PT I
(2021)
Article
Computer Science, Artificial Intelligence
Werner Mostert, Katherine M. Malan, Andries P. Engelbrecht
Summary: This study introduces a novel performance metric BFI for feature selection algorithms, which can be used for comparative analysis. The research found performance complementarity among a suite of feature selection algorithms on various real world datasets.
Article
Computer Science, Artificial Intelligence
Ryan Dieter Lang, Andries Petrus Engelbrecht
Summary: This study introduces a method to determine the minimum sample size required for robust exploratory landscape analysis measures, and utilizes self-organizing feature map to cluster a comprehensive set of benchmark functions, proposing a benchmark suite with improved coverage in single-objective, boundary-constrained problem spaces.
Article
Computer Science, Artificial Intelligence
Jonathan Mwaura, Andries P. Engelbrecht, Filipe V. Nepomuceno
Summary: This paper discusses the concepts of multimodal problems, multimodal optimisation, niche algorithms, and how diversity measures can be used to evaluate the distribution of candidate solutions and solutions of niching algorithms.
