Expert Systems with Applications最新文献

{"title":"A method for real-time detection of vessel abnormal behavior based on CNN-LSTM","authors":"Yuhao Qi ,&nbsp;Jiaxuan Yang ,&nbsp;Anzhi Bai ,&nbsp;Jiaguo Liu","doi":"10.1016/j.eswa.2025.128303","DOIUrl":"10.1016/j.eswa.2025.128303","url":null,"abstract":"<div><div>Traditional methods for detection of vessel abnormal behavior struggle to achieve real-time detection and have low detection accuracy due to the neglect of spatial characteristics of vessel trajectories. To address these issues, a method for real-time detection of vessel abnormal behavior based on Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) was proposed. Initially, leveraging the spatial feature extraction capability of CNN and the time series modeling capability of LSTM, a comprehensive CNN-LSTM model was constructed which integrates spatial–temporal features. Subsequently, an algorithm for real-time detection of vessel abnormal behavior was devised, achieving real-time detection of ship abnormal behavior by comparing denoised actual values with predicted values, which utilized rolling predictions from the model to obtain predicted values of vessel behavior features at each moment, employed Exponential Weighted Moving Average (EWMA) method to filter out noise in real-time data. Tianjin Port was chosen as the research area to validate the feasibility of the method. Experimental results indicated that the proposed method could effectively detect anomalous time points and behavior over a period of time. Compared to other approaches, the proposed method demonstrates superior performance in feature prediction. Additionally, after incorporating EWMA method, it can efficiently suppress noise interference, enhancing detection accuracy and robustness. This advancement provides substantial technical support for enhancing maritime traffic safety and vessel security management.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"288 ","pages":"Article 128303"},"PeriodicalIF":7.5,"publicationDate":"2025-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144138792","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":"Multi-order Chebyshev-based composite relation graph matching network for temporal sentence grounding in videos","authors":"GuangLi Wu,&nbsp;Xinlong Bi,&nbsp;Jing Zhang","doi":"10.1016/j.eswa.2025.127901","DOIUrl":"10.1016/j.eswa.2025.127901","url":null,"abstract":"<div><div>Temporal sentence grounding in videos (TSGV) aims to locate the semantically most matching target moment from the untrimmed video based on a query text. Existing methods have achieved certain results in cross-modal interaction by using Graph Convolutional Networks (GCNS), but the conventional graph convolutional operations have large computation and can generate a lot of redundant information. Meanwhile, the modality interactions they employ may lead to irrelevant noise and over-smoothing risks. In this paper, we proposed a Multi-order Chebyshev-based Composite Relation Graph Matching Network for Temporal Sentence Grounding in Videos. Specifically, for video feature encoder, we use the multi-order Chebyshev to approximate the graph convolution filter, which can flexibly capture long-distance dependencies in the graph structure and effectively avoid the propagation of redundant information, while ensuring computational efficiency and improving localization accuracy. For graph matching, a composite relation graph matching network is designed by constructing the connection relation discriminating module and the association relation calculating module, significantly improving the accuracy of cross-modal semantic matching. This method fully captures complex inter-modal semantic relationships, enhancing the localization accuracy of model and strengthening its capacity to represent multimodal data. We conduct comprehensive experiments on public ActivityNet Captions and TACoS datasets to prove that the method proposed in this paper is superior to state-of-the-art approaches.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"288 ","pages":"Article 127901"},"PeriodicalIF":7.5,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144124933","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":"Advancing sleep stages classification through a dual-Graphormer approach","authors":"Peilin Huang ,&nbsp;Meiyu Qiu ,&nbsp;Yi Liu ,&nbsp;Xiaomao Fan","doi":"10.1016/j.eswa.2025.128220","DOIUrl":"10.1016/j.eswa.2025.128220","url":null,"abstract":"<div><div>Sleep stages classification plays a crucial role in the diagnosis and management of sleep disorders. Recent advancements utilizing graph neural networks (GNNs) have highlighted their effectiveness in modeling the topological characteristics of sleep physiological signals, yielding promising results. Despite these advancements, a significant challenge persists in adequately addressing the spatial and temporal dependencies inherent in these signals. In this study, we present DGraphormer-SleepNet, a novel methodology designed for precise sleep stages classification, which integrates a Graphormer-based framework, a transformer variant enhancing with the structural encoding for graph data. This approach synergizes the strengths of graph neural networks and transformer architectures, employing graph-based spatial attention (GSAttn) and temporal channel attention (TCAttn) mechanisms to effectively capture both spatial and temporal dependencies within sleep physiological signals. We conduct extensive experiments using two publicly available sleep datasets, ISRUC-S1 and ISRUC-S3. The results demonstrate that DGraphormer-SleepNet significantly outperforms existing state-of-the-art methods, underscoring its potential for enhancing sleep stages classification accuracy. The source code of DGraphormer-SleepNet is available at: <span><span>https://github.com/Huang-Peilin-cn/DGraphormer-SleepNet</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"288 ","pages":"Article 128220"},"PeriodicalIF":7.5,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144170124","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":"EEG-based listened-language classification","authors":"Isaac Ariza ,&nbsp;Lorenzo J. Tardón ,&nbsp;Ana M. Barbancho ,&nbsp;Isabel Barbancho","doi":"10.1016/j.eswa.2025.128276","DOIUrl":"10.1016/j.eswa.2025.128276","url":null,"abstract":"<div><div>From an early age, individuals are continuously exposed to other languages beyond their native tongue; however, the brain’s response to these auditory stimuli remains unclear. To investigate this, an experiment was designed to record electroencephalography (EEG) signals from subjects listening to sentences in five different languages, and a specific database was built to enable performing classification tests to distinguish between different languages, and varying levels of language comprehension. By analysing the energy difference between the EEG channels to characterize these signals, different classification tests were conducted using bidirectional Long Short-Term Memory (bi-LSTM), Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) neural networks.</div><div>The main objective is the analysis of the brain’s response in two different scenarios: when the subject listens to sentences in different languages, and when the subject understands or misunderstands the meaning of a sentence. In the multi-class classification involving sentences in five different languages, the accuracy attained is 36.37 %. However, in the multi-class classification between ‘understood’/‘understood part of the meaning’/‘didn’t understand’, the accuracy attained reaches 81.36 %. The results obtained for binary classification tests of understand native language or foreign language is 89.09 %. The bi-LSTM neural network achieved the overall best performance.</div><div>These results demonstrate that the analysis of the EEG signals alone can give information regarding a person’s language comprehension level, and can be used for monitoring the learning curve of a new language or to assess comprehension in patients with conditions such as aphasia.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"288 ","pages":"Article 128276"},"PeriodicalIF":7.5,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144169658","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 teaching quality evaluation framework for blended classroom modes with multi-domain heterogeneous data integration","authors":"Yu Zhou ,&nbsp;Sai Zou ,&nbsp;Minghui Liwang ,&nbsp;Yanglong Sun ,&nbsp;Wei Ni","doi":"10.1016/j.eswa.2025.127884","DOIUrl":"10.1016/j.eswa.2025.127884","url":null,"abstract":"<div><div>The blended teaching model, combining online and traditional classroom education, has become a normalized instructional approach in the artificial intelligence (AI) era. However, effectively integrating diverse educational data, accurately evaluating student learning, and providing personalized teaching recommendations remain key challenges. To address these, this study develops an AI-driven framework for evaluating teaching quality in blended classrooms. The framework aggregates multi-source educational data from both physical and digital learning environments, enabling a comprehensive assessment of student progress. Based on the evaluation outcomes, it leverages AI to generate personalized teaching strategies aligned with each student’s learning profile. Applied on a large scale in Western China, this framework has demonstrated significant improvements in teaching quality and efficiency. By bridging educational theory and AI-driven applications, this study offers a scalable model for enhancing personalized learning experiences in global digital education.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"289 ","pages":"Article 127884"},"PeriodicalIF":7.5,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144169673","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":"Path planning algorithm of steerable flexible needle: A review","authors":"He Zhang ,&nbsp;Yan-Jiang Zhao ,&nbsp;Ye-Xin Jin ,&nbsp;Hai-Long Duan","doi":"10.1016/j.eswa.2025.128270","DOIUrl":"10.1016/j.eswa.2025.128270","url":null,"abstract":"<div><div>In minimal invasive surgery, the steerable flexible needle (SFN) has garnered widespread interest due to its high agility and low invasiveness. As one of its core technologies, the path planning algorithm is very important for the accuracy and safety of the insertion procedure. This review aims to conduct an in-depth analysis of the research trends and hotspots in the path planning of the SFN. Through a visual analysis of the researches in this field by using the bibliometric analysis tool Citespace, this review highlights the key of current researches on the path planning. In this paper, the motion mechanism and the curvature constraint of the SFN in the insertion process are firstly described. Secondly, the research methods of the path planning algorithms of the SFN are comprehensively and systematically reviewed, and are deeply discussed from four perspectives: mathematical calculation based algorithm, inverse kinematics based algorithm, sampling based algorithm and intelligence based algorithm. Furthermore, this review discusses the main factors affecting the application of the path planning algorithm at the present stage, and views the development prospects, which provides fresh ideas for researchers.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"287 ","pages":"Article 128270"},"PeriodicalIF":7.5,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144117109","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":"Multi-view syntax-semantics information bottleneck for dependency-driven relation extraction","authors":"Wei She ,&nbsp;Xiwang Li ,&nbsp;Linpu Lv ,&nbsp;Youwei Wang ,&nbsp;Honghui Dong ,&nbsp;Zhao Tian","doi":"10.1016/j.eswa.2025.128207","DOIUrl":"10.1016/j.eswa.2025.128207","url":null,"abstract":"<div><div>Dependency-driven relation extraction (DDRE) intends to improve the classification performance by exploring the syntax information contained in dependency tree. Although recent works have made impressive advances, they usually suffer severe challenges of losing sequential semantics information due to the pruning strategy. In addition, task-irrelevant information in representations such as noise or entity appositive information might be dominant in the procedure of representation learning, which interferes with the final performance of relation classification. To address these challenges, we propose a novel multi-view syntax-semantics information bottleneck (MS2IB) model, which aims at exploring supplementary information from a newly incorporated view and captures minimum sufficient information for the DDRE task. Specifically, MS2IB treats the DDRE task as a multi-view information learning procedure, where the sequential semantics information is supplemented under the guidance of representation learning with multiple views. Meanwhile, the minimum sufficient task-relevant information is extracted from the aspect of information compression and information preservation simultaneously. Finally, we formulate the objective of MS2IB as an information loss function based on the measurement of mutual information, where a new variational approach is presented to ensure its local optimum. Experiments on benchmark datasets and self-constructed dataset are conducted to show the superiority of MS2IB over the state-of-the-art models. For example, the proposed MS2IB achieves 91.42 % Precision, 89.21 % Recall and 90.28 % F1-score on the SemEval-2010 dataset. The MS2IB model further demonstrates strong generalization on domain datasets and robust performance on large-scale and noisy dataset. With the promising performance on both universal and domain datasets, the proposed model can be applied into various practical applications, such as information extraction and inference of transportation incidents/accidents.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"288 ","pages":"Article 128207"},"PeriodicalIF":7.5,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144169528","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":"Long-tailed classification based on dynamic class average loss","authors":"Do Ryun Lee,&nbsp;Chang Ouk Kim","doi":"10.1016/j.eswa.2025.128292","DOIUrl":"10.1016/j.eswa.2025.128292","url":null,"abstract":"<div><div>In real-world data distributions, class imbalance is a common issue. When training deep learning models on class-imbalanced data, the performance of classes with fewer samples tends to deteriorate. Numerous studies have addressed this problem, focusing on loss reweighting techniques based on the number of training samples per class. However, because some classes are inherently easier or harder to classify, having a larger number of samples in a particular class does not necessarily ensure lower loss or better learning for that class. Additionally, if the ratio of loss magnitudes differs substantially from the ratio of the number of training samples per class, reweighting based solely on sample size may be inappropriate. This study proposes a method to reweight losses based on dynamic class average losses rather than the number of training samples per class to address these issues. Specifically, this method evaluates the class average losses for each mini-batch, applies a nonlinear transformation to these values, and dynamically adjusts the class-wise loss weights within the loss function during training to better mitigate class imbalance. Experimental results from various types of datasets, including image and tabular data, demonstrate that the proposed method improves performance by 1%–8% across various datasets compared to existing methods.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"288 ","pages":"Article 128292"},"PeriodicalIF":7.5,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144154946","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":"DKSCNN: Deep Kronecker Siamese Convolutional Neural Network enabled speaker identification","authors":"Karthikeyan Chinnasamy ,&nbsp;Rajesh Kumar Thevasigamani ,&nbsp;Rajiv Vincent ,&nbsp;Sam Kumar Gopalsamy Venkatesan ,&nbsp;Deepa Thilak Kanniyappan ,&nbsp;Kalaiselvi Kaliannan","doi":"10.1016/j.eswa.2025.127946","DOIUrl":"10.1016/j.eswa.2025.127946","url":null,"abstract":"<div><div>Speaker identification refers to the process of discerning between different voices within audio recordings or streams. Various factors contribute to the complexity of this task, including differences in frameworks, overlapping of sound events, and multiple sound sources present during recording. These aspects significantly complicate the process of speaker identification. To overcome such complexity, a hybrid Deep Kronecker Siamese Convolutional Neural Network (DKSCNN) method is proposed as a solution for performing speaker identification. Initially, the speech signals are collected from the VoxCeleb dataset and it is fed as an input to the pre-processing step and it is performed by the Gaussian distribution-based method. After preprocessing, feature extraction is done to extract features like spectral centroid, pitch chroma, spectral skewness, Power Spectral Density (PSD), Mel-Scale Frequency Cepstral Coefficients (MFCC), logarithmic band power, Hjorth parameters, and tonal power ratio. Based on the extracted features, the speaker identification is done using hybrid DKSCNN, which is the combination of Deep Kronecker Network (DKN) and Siamese Convolutional Neural Network (SCNN). The proposed speaker identification model attained a better accuracy of 90.682%, a True Positive Rate (TPR) rate of 91.362% and a False Positive Rate (FPR) rate of 0.086. The DKSCNN model significantly improves the accuracy and reliability of speaker identification, achieving remarkable performance metrics. This research contributes to the enhancement of speaker identification technology and addresses the problems of real-world audio environments. Also, this approach ensures that speaker identification can be applied across diverse applications.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"288 ","pages":"Article 127946"},"PeriodicalIF":7.5,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144154940","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 self-supervised deep learning framework for seismic facies segmentation","authors":"Ming Li ,&nbsp;Xue-song Yan ,&nbsp;Qing-hua Wu","doi":"10.1016/j.eswa.2025.128290","DOIUrl":"10.1016/j.eswa.2025.128290","url":null,"abstract":"<div><div>Seismic facies classification plays a vital role in subsurface exploration and geological interpretation, but the dependence on manually labeled seismic data can hinder scalability and efficiency. In this paper, we propose SSFS (Self-supervised Seismic Facies Segmentation), a novel framework that combines deep clustering with self-supervised learning to achieve seismic facies segmentation without the need for labeled data. Traditional methods rely on costly manual labeling or simplistic clustering, while supervised deep learning struggles with limited labeled data. SSFS addresses these limitations by deep clustering and self-supervised learning. The SSFS framework consists of three key phases: (1) Deep Clustering, where seismic data is sliced into overlapping patches of varying sizes and features are extracted using an attention-based auto-encoder, followed by clustering the features in latent space using K-means; (2) Cluster Merging, where initial clusters are iteratively merged based on cosine similarity to refine the clustering results; and (3) Facies Segmentation, where the merged clusters serve as pseudo-labels for training a segmentation model to refine the facies classification. On the Netherlands F3 dataset, SSFS achieves an average mean Intersection over Union (mIoU) of 83.76% and average pixel accuracy (PA) of 88.43% on four test seismic profiles. Moreover, we show that SSFS can enhance supervised seismic facies analysis by initializing popular segmentation models with the pre-trained weights from the SSFS framework. Quantitative validations across two datasets demonstrate that SSFS-initialized models consistently improve class accuracy (CA) by 2–5% compared to models trained from scratch for minority facies classes. The effectiveness of SSFS is quantitatively validated through several metrics, highlighting its potential for both self-supervised and supervised seismic facies classification tasks. Our results suggest that SSFS provides a promising solution for seismic facies analysis, especially in scenarios with limited labeled data, and has the potential to significantly improve seismic interpretation workflows in practice.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"288 ","pages":"Article 128290"},"PeriodicalIF":7.5,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144138795","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}