Yifan Liu , Hongmei Ma , Donglin Pan , Yi Liu , Wenlei Chai , Zhenpeng Liu
{"title":"HierFLMC: Efficient hierarchical federated learning based on soft clustering model compression","authors":"Yifan Liu , Hongmei Ma , Donglin Pan , Yi Liu , Wenlei Chai , Zhenpeng Liu","doi":"10.1016/j.eswa.2025.130022","DOIUrl":"10.1016/j.eswa.2025.130022","url":null,"abstract":"<div><div>Federated Learning (FL) is an advanced distributed machine learning paradigm that enables clients to collaboratively train a shared neural network model using local datasets, transmitting model parameters instead of raw data. However, in many FL systems, the frequent exchange of model parameters between clients and remote cloud servers leads to significant communication overhead. As the model size increases, existing FL methods incur substantial communication costs. To address this bottleneck, this paper proposes a novel hierarchical federated learning model compression scheme (HierFLMC). This scheme integrates model compression techniques within a hierarchical framework, significantly enhancing communication efficiency between edge devices and cloud servers. In addition, an innovative preliminary soft clustering model update compression algorithm (SCMUC) is proposed. The SCMUC algorithm utilizes the K-means method for initial clustering, effectively reducing the computational complexity associated with traditional soft clustering methods. We validate the proposed scheme using the CIFAR-10 and FEMNIST datasets, demonstrating a 2 % improvement in model accuracy and an 11 % reduction in communication time compared to MUCSC. Experimental results indicate that this approach not only achieves a favorable compression rate but also substantially improves communication efficiency.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130022"},"PeriodicalIF":7.5,"publicationDate":"2025-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364912","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ruixin Wang , Zhenghong Wu , Jiangfeng Fu , Han Zhang , Haidong Shao
{"title":"Universal source-free domain adaptation method with feature decomposition for machinery fault diagnosis","authors":"Ruixin Wang , Zhenghong Wu , Jiangfeng Fu , Han Zhang , Haidong Shao","doi":"10.1016/j.eswa.2025.129986","DOIUrl":"10.1016/j.eswa.2025.129986","url":null,"abstract":"<div><div>The rapid advancement of domain adaptation algorithms has significantly accelerated the deployment of intelligent diagnostic technologies. However, existing domain adaptation methods predominantly focus on closed-set fault diagnosis and rarely address data privacy concerns, limiting their applicability in industrial settings. To this end, a universal source-free domain adaptation method is proposed. Initially, a source model is pre-trained using labeled source data. This pre-trained model then processes the target data to decompose the target features into common class components and unknown class components, while simultaneously generating target prototypes and source anchors. Subsequently, the distribution of the unknown class components is estimated using a Gaussian mixture model with two components. Finally, a confidence estimation strategy is developed to derive instance-level decision boundaries by evaluating the distance between target prototypes and source anchors, thereby completing the classification task. Experimental results on gearbox and rolling bearing datasets demonstrate that our approach excels in handling fault diagnosis under varying conditions while ensuring data privacy.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 129986"},"PeriodicalIF":7.5,"publicationDate":"2025-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145334064","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yuanji Shen , Jun Dai , Mingxian Wang , Gonzalo R. Arce
{"title":"Stock price trend forecasting based on multi-channel complementary network with CEEMDAN decomposition and transformer residual prediction","authors":"Yuanji Shen , Jun Dai , Mingxian Wang , Gonzalo R. Arce","doi":"10.1016/j.eswa.2025.130028","DOIUrl":"10.1016/j.eswa.2025.130028","url":null,"abstract":"<div><div>Forecasting stock market movement provides important investment signals, but the presence of a large amount of noise in financial time series (FTS) data poses significant challenges to prediction accuracy. This paper proposes a novel multi-channel complementary network with data decomposition and a fusion strategy to effectively improve the prediction accuracy and robustness of stock price movement. The model first applies the algorithm of Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) to decompose the historical stock data sequence into a set of Intrinsic Mode Functions (IMFs), representing different frequency components of the original FTS data. Then, each IMF combined with six key indicators (i.e., open price, high price, low price, trading volume, price-earnings ratio and price-to-book ratio) are processed by an independent long short-term memory (LSTM) module to make a parallel prediction. To further enhance the forecasting accuracy, a transformer-based residual term prediction model is incorporated, serving as the complementary branch to the LSTM modules. Subsequently, the outputs from all network branches are fused together to obtain the final prediction result. A set of numerical experiments on different stock index datasets are conducted to verify the superiority of the proposed model in terms of average forecasting accuracy compared with other benchmark models. In addition, the effectiveness of different sub-modules in the proposed framework is proved by the ablation experiments.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130028"},"PeriodicalIF":7.5,"publicationDate":"2025-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364853","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Color image encryption scheme based on 5D fractional-order complex chaotic system and eight-base DNA cubes","authors":"Liming Wu , Zijian Tian , Wei Chen","doi":"10.1016/j.eswa.2025.129950","DOIUrl":"10.1016/j.eswa.2025.129950","url":null,"abstract":"<div><div>With the rapid development of information technology, the secure transmission of color images has become an urgent problem. In order to overcome this problem, this paper proposes a high-security color image encryption algorithm. First, a five-dimensional fractional-order complex chaotic system is designed, and its dynamics are comprehensively analyzed from the perspectives of Lyapunov exponent spectrum, bifurcation diagram, and phase diagram. The results prove that the system has good chaotic characteristics. Second, a dynamic Arnold scrambling algorithm is designed, which dynamically generates scrambling parameters by calculating the pixel values of each channel of an image. Then, combined the eight-base DNA coding scheme with a three-dimensional cube structure, a space rotation-based scrambling method and a diffusion method integrating DNA mutation and crossover operations were designed under chaotic sequence control, thereby enhancing the nonlinearity and security of the encryption system. Finally, simulation experiments and security analysis indicate that the proposed color image encryption scheme provides high security and strong key sensitivity. It effectively scrambles pixel distributions, demonstrating robustness against statistical, noise, and cropping attacks, while maintaining efficient computational performance. These results demonstrate its effectiveness for secure digital image transmission.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 129950"},"PeriodicalIF":7.5,"publicationDate":"2025-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364839","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"A prototype-based framework for open-set heterogeneous federated face recognition","authors":"Taeyong Kim, Jungyun Kim, Andrew Beng Jin Teoh","doi":"10.1016/j.eswa.2025.130020","DOIUrl":"10.1016/j.eswa.2025.130020","url":null,"abstract":"<div><div>Federated learning enables privacy-preserving face recognition by keeping raw images on users’ devices, addressing challenges of non-IID identity distributions and heterogeneous model architectures. To tackle these, we propose FedPFR, a prototype-based, architecture-agnostic framework for open-set Heterogeneous Federated Face Recognition (HtFFR). Each client computes identity prototypes-mean feature embeddings of local classes-that are uploaded to the server and redistributed without averaging, while local models are trained with a hybrid loss combining CosFace and a novel prototype-anchor contrastive (PAC) loss. This design preserves semantic integrity by reducing the distance between local embeddings and their global prototypes and enlarging inter-class separation. We provide a mathematical convergence analysis proving that FedPFR converges to a stationary point under appropriate learning conditions. Extensive experiments on the IJB-C benchmark with 20 heterogeneous clients show that FedPFR achieves strong verification performance, with 23.39 % TAR at FAR=1e-6, outperforming local-only training (21.70 %) and prior heterogeneous FL baselines. Furthermore, our cost analysis quantifies the computation, communication, and storage overhead, confirming the framework’s scalability and practicality. Ablation studies and clustering analyses further demonstrate that FedPFR produces compact and discriminative embeddings, highlighting its robustness as a resource-efficient solution for real-world open-set federated face recognition.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130020"},"PeriodicalIF":7.5,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364852","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Operating performance assessment and fault diagnosis of electro-hydraulic servo valves using Trans-SN-CGAN with Gaussian and non-Gaussian information fusion","authors":"Hanwen Zhang , Wenxiao Yin , Chuanfang Zhang , Qiang Min , Ruihua Jiao , Kaixiang Peng","doi":"10.1016/j.eswa.2025.129940","DOIUrl":"10.1016/j.eswa.2025.129940","url":null,"abstract":"<div><div>The electro-hydraulic servo valve (EHSV) is a critical component in hydraulic systems, particularly in high-precision applications such as aircraft braking systems. Traditional operational performance assessment (OPA) and fault diagnosis (FD) methods often struggle to capture the complex spatiotemporal relationships and non-Gaussian characteristics inherent in EHSV data, especially when dealing with imbalanced datasets. To address these challenges, this paper presents a Transformer-based spectral normalization conditional generative adversarial network (Trans-SN-CGAN) for monitoring the EHSV. Trans-SN-CGAN integrates both Gaussian and non-Gaussian information through kernel principal component analysis (KPCA) and kernel independent component analysis (KICA). It further transforms time-series data into 2D images using Gramian angular fields (GAF), preserving essential spatiotemporal features. Additionally, spectral normalization (SN) is incorporated into the conditional generative adversarial network (CGAN) to enhance training stability and generate high-quality synthetic data samples. The Transformer architecture is utilized to enhance the model’s capacity to extract discriminative feature representations by capturing long-range dependencies in sequential data. Experimental results demonstrate exceptional performance, with OPA accuracy of 97.6 % and FD accuracy of 97.2 %, showcasing the framework’s robustness and reliability for improving EHSV efficiency and dependability.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 129940"},"PeriodicalIF":7.5,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364914","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Meifang Zhang , Jing Bi , Haitao Yuan , Ziqi Wang , Jia Zhang , Rajkumar Buyya
{"title":"An enhanced hybrid deep neural network method for adjusted industrial time series prediction with variable operating states","authors":"Meifang Zhang , Jing Bi , Haitao Yuan , Ziqi Wang , Jia Zhang , Rajkumar Buyya","doi":"10.1016/j.eswa.2025.130029","DOIUrl":"10.1016/j.eswa.2025.130029","url":null,"abstract":"<div><div>In industrial production, dynamic nature of working conditions and reliance on manual judgment introduces significant hurdles for accurate prediction models. Despite commendable performance of contemporary Deep Learning techniques in time series prediction (TSP), they frequently overlook crucial impact of human intervention. Moreover, the subjective nature of operational condition labeling and the scarcity of comprehensive experimental datasets further hinder the efficacy of predictive systems. This work proposes an <u>E</u>nhanced <u>H</u>ybrid <u>D</u>eep <u>N</u>eural <u>N</u>etwork (EH-DNN) framework to tackle these issues. It achieves robust classification and prediction of working conditions by integrating the multi-dimensional features of set values and observation time series. The data preprocessing phase encompasses feature extraction and feature fusion, ensuring the model acquires the essential information intrinsic to the production process. A novel two-step prediction methodology is employed during the training phase, incorporating pre-classification to enhance TSP, achieving an accuracy of 94%. EH-DNN mirrors intricate dynamics of industrial production and aligns seamlessly with real-world application scenarios, demonstrating substantial practical utility. By integrating this methodology, the industrial sector can anticipate a significant leap in automation levels and production efficiency, bridging the gap between theoretical models and practical implementation.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130029"},"PeriodicalIF":7.5,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364915","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"An acoustic sensing system for noise monitoring and source identification using transfer learning","authors":"Dolvara Gunatilaka, Wudhichart Sawangphol, Thanakorn Charoenritthitham, Thanawat Kanjanapoo, Teerapat Burasotikul, Kittikawin Pongprasit","doi":"10.1016/j.eswa.2025.130014","DOIUrl":"10.1016/j.eswa.2025.130014","url":null,"abstract":"<div><div>Increasing noise pollution in urban areas underscores the need for an autonomous system to monitor and control noise. Beyond detecting noise levels, identifying noise sources further improves noise management. This work presents a scalable IoT-based sensing platform for smart environment applications. The system integrates low-cost devices for acoustic measurement, edge devices to enable noise source identification, a back-end infrastructure crucial for efficient acoustic data and device management, and a web-based application facilitating noise data visualization. Our study explores three feature extraction techniques and eight Convolutional Neural Network (CNN)-based pre-trained models for noise classification on the resource-constrained Raspberry Pi platform and compares their performance. Leveraging pre-trained models helps speed up the model development process. UrbanSound8k, ESC-50 datasets, and audio data collected with our low-cost microphone are used for model development and validation. The evaluation results show that our hierarchical model, utilizing the Mel Spectrogram feature extraction method and a MobileNet model, achieves the highest accuracy of 90.18 %. Furthermore, we deploy the system and assess its performance. Our system can reliably transmit audio data with an average delay of 0.37 s, and the Raspberry Pi can perform feature extraction and classification within an average of 2.5 s. Hence, our solution offers a comprehensive and cost-effective solution to enhance noise management and control.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130014"},"PeriodicalIF":7.5,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364918","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Xiaoyang Zheng , Zejiang Yu , Lei Chen , Zijian Lei , Zhixia Feng
{"title":"Optimal Legendre multiwavelet frequency band-based an improved adaptive denoising algorithm for mechanical fault diagnosis under complex conditions","authors":"Xiaoyang Zheng , Zejiang Yu , Lei Chen , Zijian Lei , Zhixia Feng","doi":"10.1016/j.eswa.2025.130025","DOIUrl":"10.1016/j.eswa.2025.130025","url":null,"abstract":"<div><div>Conventional fault diagnosis methods face significant challenges in real-world engineering scenarios, including heavy background noise, wrong labels, and limited fault samples. To address these challenges, this paper proposes an improved adaptive denoising algorithm combining Legendre multiwavelet with genetic algorithm (IAD-LWGA). This novel method devises a modified threshold estimation algorithm on each LW frequency band optimized by GA, resulting in effectively suppressing noise while preserving fault-related features. The efficacy of the proposed approach is first validated on a real-world air conditioner external unit dataset. Subsequently, the extracted optimal feature combinations are directly transferred to PHM 2009 gearbox compound fault diagnosis dataset, demonstrating strong cross-domain generalization ability with minimal requirement for domain-specific expertise. Extensive experiments show that the proposed approach consistently outperforms state-of-the-art models, attaining 100 % accuracy for both datasets under normal conditions, while reaching 100 %, 94.75 %, 93.57 % accuracies with –10 dB noise, label noise ratio 0.05, faulty samples 10 for Dataset 1, and achieving 99.83 %, 95.33 %, 90.80 % accuracies with 6 dB noise, label noise ratio 0.05, limited faulty samples 10 for Dataset 2, respectively. This work presents a practical and effective strategy for fault diagnosis in complex industrial environments, enhancing predictive maintenance capabilities of expert and intelligent systems.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"299 ","pages":"Article 130025"},"PeriodicalIF":7.5,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364913","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}