Exploring the potential of combining Mel spectrograms with neural networks optimized by the modified crayfish optimization algorithm for acoustic speed violation identification DOI
Marko Stankovic, Luka Jovanovic,

Aleksandra Bozovic

et al.

International Journal of Hybrid Intelligent Systems, Journal Year: 2024, Volume and Issue: 20(2), P. 119 - 143

Published: May 31, 2024

Enforcing vehicle speed limits is paramount for road safety. This paper pioneers an innovative approach by synergizing signal processing and Convolutional Neural Networks (CNNs) to detect speeding violations, addressing a critical aspect of traffic management. While traditional methods have shown efficacy, the potential synergy AI techniques remains largely unexplored. We bridge this gap harnessing Mel spectrograms extracted from recordings, representing intricate audio features. These serve as inputs tailored CNN architecture, meticulously designed pattern recognition in speeding-related cues. An altered variant crayfish optimization algorithm (COA) was employed tune model. Our methodology aims discriminate between normal driving sounds instances limit breaches. Notably absent previous literature, our fusion method yields promising initial results, demonstrating its accurately identify violations. contribution not only enhances safety management but also provides pioneering framework integrating ways, with implications extending broader analysis domains.

Language: Английский

Intrusion detection in metaverse environment internet of things systems by metaheuristics tuned two level framework DOI Creative Commons
Miloš Antonijević, Miodrag Živković, Milica Djurić-Jovičić

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Jan. 28, 2025

Internet of Things (IoT) is one the most important emerging technologies that supports Metaverse integrating process, by enabling smooth data transfer among physical and virtual domains. Integrating sensor devices, wearables, smart gadgets into environment enables IoT to deepen interactions enhance immersion, both crucial for a completely integrated, data-driven Metaverse. Nevertheless, because devices are often built with minimal hardware connected Internet, they highly susceptible different types cyberattacks, presenting significant security problem maintaining secure infrastructure. Conventional techniques have difficulty countering these evolving threats, highlighting need adaptive solutions powered artificial intelligence (AI). This work seeks improve trust in edge integrated study revolves around hybrid framework combines convolutional neural networks (CNN) machine learning (ML) classifying models, like categorical boosting (CatBoost) light gradient-boosting (LightGBM), further optimized through metaheuristics optimizers leveraged performance. A two-leveled architecture was designed manage intricate data, detection classification attacks within networks. thorough analysis utilizing real-world network dataset validates proposed architecture's efficacy identification specific variants malevolent assaults, classic multi-class challenge. Three experiments were executed open public, where top models attained supreme accuracy 99.83% classification. Additionally, explainable AI methods offered valuable supplementary insights model's decision-making supporting future collection efforts enhancing systems.

Language: Английский

Citations

4
An optimized ensemble grey wolf-based pipeline for monkeypox diagnosis DOI Creative Commons
Ahmed I. Saleh, Asmaa H. Rabie, Shimaa E. ElSayyad

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Jan. 30, 2025

As the world recovered from coronavirus, emergence of monkeypox virus signaled a potential new pandemic, highlighting need for faster and more efficient diagnostic methods. This study introduces hybrid architecture automatic diagnosis by leveraging modified grey wolf optimization model effective feature selection weighting. Additionally, system uses an ensemble classifiers, incorporating confusion based voting scheme to combine salient data features. Evaluation on public sets, at various training samples percentages, showed that proposed strategy achieves promising performance. Namely, yielded overall accuracy 98.91% with testing run time 5.5 seconds, while using machine classifiers small number hyper-parameters. Additional experimental comparison reveals superior performance over literature approaches metrics. Statistical analysis also confirmed AMDS outperformed other models after running 50 times. Finally, generalizability is evaluated its external sets COVID-19. Our achieved 98.00% 99.00% COVID respectively.

Language: Английский

Citations

1
Two-tier deep and machine learning approach optimized by adaptive multi-population firefly algorithm for software defects prediction DOI
John Philipose Villoth, Miodrag Živković, Tamara Živković

et al.

Neurocomputing, Journal Year: 2025, Volume and Issue: unknown, P. 129695 - 129695

Published: Feb. 1, 2025

Language: Английский

Citations

1
Intrusion detection using metaheuristic optimization within IoT/IIoT systems and software of autonomous vehicles DOI Creative Commons
Pavle Dakić, Miodrag Živković, Luka Jovanovic

et al.

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Oct. 2, 2024

Language: Английский

Citations

8
Computer-Vision Unmanned Aerial Vehicle Detection System Using YOLOv8 Architectures DOI Open Access
Aleksandar Petrović, Nebojša Bačanin, Luka Jovanovic

et al.

International Journal of Robotics and Automation Technology, Journal Year: 2024, Volume and Issue: 11, P. 1 - 12

Published: May 22, 2024

Abstract: This work aims to test the performance of you only look once version 8 (YOLOv8) model for problem drone detection. Drones are very slightly regulated and standards need be established. With a robust system detecting drones possibilities regulating their usage becoming realistic. Five different sizes were tested determine best architecture size this problem. The results indicate high across all models that each is used specific case. Smaller suited lightweight approaches where some false identification tolerable, while largest with stationary systems require precision.

Language: Английский

Citations

6
Particle swarm optimization tuned multi-headed long short-term memory networks approach for fuel prices forecasting DOI
Andjela Jovanovic, Luka Jovanovic, Miodrag Živković

et al.

Journal of Network and Computer Applications, Journal Year: 2024, Volume and Issue: 233, P. 104048 - 104048

Published: Nov. 7, 2024

Language: Английский

Citations

6
Exploration and comparison of the effectiveness of swarm intelligence algorithm in early identification of cardiovascular disease DOI Creative Commons

Tiantian Bai,

Mengru Xu,

Taotao Zhang

et al.

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 7, 2025

Due to the aging of global population and lifestyle changes, cardiovascular disease has become leading cause death worldwide, causing serious public health problems economic pressures. Early accurate prediction is crucial reducing morbidity mortality, but traditional methods often lack robustness. This study focuses on integrating swarm intelligence feature selection algorithms (including whale optimization algorithm, cuckoo search flower pollination Harris hawk particle genetic algorithm) with machine learning technology improve early diagnosis disease. systematically evaluated performance each algorithm under different sizes, specifically by comparing their average running time objective function values identify optimal subset. Subsequently, selected subsets were integrated into ten classification models, a comprehensive weighted evaluation was performed based accuracy, precision, recall, F1 score, AUC value model determine configuration. The results showed that random forest, extreme gradient boosting, adaptive boosting k-nearest neighbor models best combined dataset (weighted score 1), where set consisted 9 key features when size 25; while Framingham dataset, 0.92), its derived from 10 50. this show can effectively screen informative sets, significantly provide strong support for diseases.

Language: Английский

Citations

0
Modeling and Multi-Objective Optimization of Transcutaneous Energy Transmission Coils Based on Artificial Intelligence DOI Open Access
Ying Mao, Xiao Li

Electronics, Journal Year: 2025, Volume and Issue: 14(7), P. 1381 - 1381

Published: March 29, 2025

This paper proposes a machine learning-based modeling and multi-objective optimization method for transcutaneous energy transfer coils to address the problem that current with single-objective design methods have difficulty achieving optimal solutions. From design, whole coil process is covered by this approach. approach models using Extreme Learning Machine, Gray Wolf Optimization algorithm used tune Machine’s parameters in order increase accuracy. The Non-Dominated Sorting Whale utilized of coils, which based on established model. Using planar helical applied artificial detrusors as an example, verification analysis was conducted, final results were demonstrated. indicate significantly outperforms comparison algorithms tuning Machine model, it exhibits good convergence ability stability. prediction model comparative terms evaluation metrics predicting three outputs (transmission efficiency, coupling coefficient, secondary diameter), demonstrating excellent performance. performs well showing results. Pareto solutions obtained errors 3.03%, 0.1%, 1.7% transmission diameter, respectively, when compared simulation experimental calculations. small validate correctness effectiveness proposed method.

Language: Английский

Citations

0
Machine learning modeling and multi objective optimization of artificial detrusor DOI Creative Commons
Yin Mao, Xiao Li

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: April 22, 2025

To address the problem of obtaining optimal design parameters for existing artificial detrusors using single-objective optimization methods, this research proposed a machine learning-based detrusor modeling and multi-objective approach, which includes thorough process from to optimization. Extreme learning was used model in suggested order increase accuracy, multi-strategy modified crayfish algorithm tuning extreme machine's put forth research. The grey wolf utilized optimize based on model. In validate an driven by shape memory spring finally built as experimental platform. results show that improved paper can effectively avoid defects original algorithm, its performance convergence ability are better than comparison algorithm. With root mean square error 1.51E-02, coefficient determination 9.81E-01, absolute 1.32E-02, percentage 1.66E-01, established predicts spring-driven detrusor's emptying rate. It also temperature increment with 8.47E-01, 5.81E-01, 7.23E-02. These predictions superior prediction model, indicating good predictive stability. Additionally, demonstrates outstanding uncertainty reliability analysis, thereby further confirming comprehensive performance. optimized computed values rate increment, well measurement values, have errors 7.8% 11.8%, respectively, satisfy engineering specifications. our method exhibits significant enhancements over designs. Specifically, achieves approximately 20% 62% reduction successfully balancing urinary efficiency mitigated risks thermal tissue injury.

Language: Английский

Citations

0
A Robust Automated Cervical Cancer Detection System Using Elephant Herding Optimized MCNN DOI

S. Maheswari,

C. N. Marimuthu,

Xavier Fernando

et al.

IGI Global eBooks, Journal Year: 2025, Volume and Issue: unknown, P. 213 - 234

Published: May 2, 2025

Cervical cancer is a leading cause of cancer-related deaths among women, and early detection crucial for improving survival rates. This research proposes an automated system classifying cervical using medical images. The starts with image preprocessing, where images are resized noise removed Median Filter. Segmentation performed K-Means Clustering to isolate cancerous regions. Local Binary Pattern (LBP) technique applied feature extraction, capturing texture patterns distinguish normal from abnormal tissues. Classification achieved Modified Convolutional Neural Network (MCNN), optimization through the Elephant Herding Optimization (EHO) algorithm fine-tune model's parameters. approach aims assist healthcare professionals in diagnosing more efficiently accurately, patient outcomes. can provide rapid, reliable results, enabling timely treatment potentially reducing global burden cancer.

Language: Английский

Citations

0