Expert Systems with Applications最新文献

{"title":"Short-term high-speed rail passenger flow forecasting integrated extended empirical mode decomposition with multivariate and bidirectional support vector machine","authors":"Xueyi Guan ,&nbsp;Michael Z.F. Li ,&nbsp;Jin Qin ,&nbsp;Chengna Wang","doi":"10.1016/j.eswa.2025.129870","DOIUrl":"10.1016/j.eswa.2025.129870","url":null,"abstract":"<div><div>High-speed rail (HSR) short-term passenger flow forecasting is of great significance for dynamically adjusting operation plans and optimizing transportation resource allocation. For this reason, this paper proposes an innovative complete ensemble empirical mode decomposition with adaptive noise integrated with multivariate and bidirectional support vector machine (CEEMDAN-MBSVM) method with four key steps. First, we analyze the correlations between multiple origin–destination (OD) passenger flows and select strongly correlated ODs incorporated with their opposite OD for joint bidirectional forecasting. Second, we decompose the original passenger flow time series by using period division technique of CEEMDAN, which yield multiple intrinsic mode functions (IMFs) and a residual trend term (RES). Then we apply MBSVM to predict the IMFs of each OD and use trend extrapolation to forecast the RES. Finally, we reconstruct the predicted IMFs and RES to obtain the final bidirectional HSR OD daily passenger flows. Subsequently, we conduct a comprehensive validation exercise and significance testing, using real data from Beijing-Shanghai HSR Line, against seven prediction methods. In particular, for five selected ODs, benchmarking against EEMD-MSVM method, the best performer among the six existing models, our model reduces the minimum mean absolute percentage error (MAPE) by 1.30 % to 4.97 % and benchmarking against ARIMA model, the worst performer among the six existing models, our model reduces the MAPE by 11.57 % to 22.72 %. This research has clearly demonstrated the value of leveraging bidirectional OD data on improving short-term passenger flow forecasting.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129870"},"PeriodicalIF":7.5,"publicationDate":"2025-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158462","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 Dynamic Tri-Stage Framework with Neural Network-Assisted Search for Constrained Multi-objective Optimization","authors":"Qianlong Dang ,&nbsp;Xinkang Hong ,&nbsp;Xianpeng Sun","doi":"10.1016/j.eswa.2025.129761","DOIUrl":"10.1016/j.eswa.2025.129761","url":null,"abstract":"<div><div>Constrained multi-objective optimization problems involve the optimization of multiple objective functions and the satisfaction of different constraints, which poses a challenge for algorithms to achieve a good balance between convergence and diversity. However, indiscriminately enhancing diversity can hinder convergence, while solely focusing on convergence may impair the exploration of the objective space, especially when the current stage is not well-defined. To address this issue, we propose a three-stage multi-task framework for constrained multi-objective optimization with dynamically switchable stages. This framework introduces two auxiliary tasks: one that operates during the exploration and transition stages to accelerate convergence towards the boundary of the infeasible regions and assist the population in crossing it, and another that operates in the final convergence stage to guide the population towards the constrained Pareto front. Moreover, a stage detection method is proposed, which evaluates the current stage to determine the appropriate evolutionary direction for the population, thus enabling dynamic stage transitions. In addition, a neural network-assisted search operator is designed for the auxiliary task during the transition stage, which learns the optimal offspring generation process. This operator enhances the ability of the auxiliary population to cross the infeasible regions. Finally, the performance of the proposed algorithm is superior and competitive on three test suites and six real-world engineering problems compared to seven state-of-the-art algorithms.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129761"},"PeriodicalIF":7.5,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145109324","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":"OSATG-GPT: Instruction-tuning large language models with open-source atomic tasks in Github","authors":"Fanyu Han ,&nbsp;Li Ma ,&nbsp;Fenglin Bi ,&nbsp;Yantong Wang ,&nbsp;Mingdong You ,&nbsp;Wei Wang ,&nbsp;Jiaheng Peng ,&nbsp;Xiaoya Xia","doi":"10.1016/j.eswa.2025.129819","DOIUrl":"10.1016/j.eswa.2025.129819","url":null,"abstract":"<div><div>Across numerous application scenarios in Natural Language Processing (NLP), Large Language Models (LLMs) have demonstrated exceptional capabilities in text comprehension and generation. These models exhibit significant potential across various interdisciplinary fields. However, their effectiveness is somewhat constrained by the unique characteristics of the open-source ecosystem. Developing an LLM with generalization capabilities across datasets and tasks, specifically tailored for the open-source ecosystem, is an urgent research need. To address this challenge, this paper introduces open-source atomic tasks, which are defined as intermediate tasks essential for solving complex objectives. These tasks are designed through strategies such as simplification, reversal, decomposition, and composition, enabling models to gradually acquire domain knowledge and understand task interdependencies. By integrating public resources with open-source atomic tasks, we construct OSE-Instruct–an instruction dataset for the open-source ecosystem. We first unify open-source atomic tasks within an instruction-tuning paradigm that reflects real-world developer behavior, and develop OSATG-GPT at various parameter scales by fine-tuning the BLOOMZ backbone model on OSE-Instruct. This enables the model to learn fine-grained developer actions and the underlying task dependencies. Extensive experiments validate the effectiveness of OSATG-GPT compared to other advanced LLMs with larger parameter scales, and highlight its advantages over GPT-4 in specific and complex open-source collaboration tasks.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129819"},"PeriodicalIF":7.5,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221477","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":"SUNRISE: multi-agent reinforcement learning via neighbors’ observations under fully noisy environments","authors":"Kaiyu Wang ,&nbsp;Bohao Qu ,&nbsp;Menglin Zhang ,&nbsp;Xianchang Wang ,&nbsp;Ximing Li","doi":"10.1016/j.eswa.2025.129781","DOIUrl":"10.1016/j.eswa.2025.129781","url":null,"abstract":"<div><div>Multi-agent reinforcement learning (MARL) methodologies have achieved notable advancements across diverse domains. Despite these successes, the susceptibility of neural networks to perturbed data and the ubiquity of external attacks in real-world settings, such as sensor noise, pose challenges for MARL approaches. The pivotal issue revolves around the effective transfer of policies learned in idealized simulation environments to the complexities inherent in real-world scenarios. More precisely, when agents are unable to obtain any accurate observations of the external environment throughout the entire policy learning process, the MARL methods cannot learn effective policies. In addressing this issue, we propose a methodology wherein noisy observations from neighboring agents are utilized, with an agent’s own noisy observations serving as surrogate ground truth. This approach facilitates the learning of effective policies by MARL methods in environments characterized by pervasive noise. We design a denoising representation network to filter out the principal state information from environment data characterized by noise to mitigate the adverse effects of noise on the process of policy learning. Then, we integrate the denoising representation network with classic MARL methodologies to learn effective policies within environments characterized by pervasive noise. A series of exhaustive experimental results demonstrate the efficacy of our approach in attenuating the impact of external attacks on the optimization parameters of neural networks during the policy-learning process. Moreover, our methodology exhibits compatibility with classic MARL methods, allowing for the learning of effective policies.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129781"},"PeriodicalIF":7.5,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221479","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":"CR-GAC: Cross-modal Recombination via Graph-Attention Collaborative Optimization for multimodal sentiment analysis","authors":"Haoran Chen ,&nbsp;Jiapeng Liu ,&nbsp;Zuhe Li ,&nbsp;Yushan Pan ,&nbsp;Hongwei Tao ,&nbsp;Huaiguang Wu ,&nbsp;Yunyang Wang ,&nbsp;Chenguang Yang","doi":"10.1016/j.eswa.2025.129805","DOIUrl":"10.1016/j.eswa.2025.129805","url":null,"abstract":"<div><div>Multimodal sentiment analysis integrates linguistic, audio, and visual modalities for predicting human emotional states. However, current algorithms encounter three challenges: limitations in adjacency matrix modeling, noise interference and modality imbalances in cross-modal attention, and inefficient cross-modal feature alignment. To address these, we propose the <strong>C</strong>ross-modal <strong>R</strong>ecombination via <strong>G</strong>raph-<strong>A</strong>ttention <strong>C</strong>ollaborative Optimization (CR-GAC) by unifying graph and sequence learning in a collaborative framework. Specifically, we <strong>first</strong> design the modality-adaptive <strong>M</strong>ultimodal <strong>G</strong>raph <strong>C</strong>onstruction (MGC) to tackle the first challenge. For the linguistic modality, a local sparse graph based on a K-Nearest Neighbors-Radial Basis Function kernel is designed to preserve fine-grained semantics; for the audio and visual modalities, a low-rank representation method combined with nuclear norm regularization is designed to capture latent cross-sample structures via singular value decomposition, while suppressing noise interference. Modalities that have been processed are then input into graph attention networks to achieve higher-order feature aggregation. <strong>Next,</strong> we construct the <strong>L</strong>anguage-guided <strong>H</strong>ierarchical <strong>C</strong>ross-modal <strong>I</strong>nteraction (LHCI) to tackle the second challenge, which leverages bidirectional cross-modal attention and multi-level Transformer blocks to hierarchically enhance feature representations. <strong>Subsequently,</strong> the <strong>H</strong>igh-level <strong>M</strong>ultimodal <strong>F</strong>eature <strong>C</strong>ontainer (HMFC) iteratively accumulates multi-grained semantics, providing a high-level feature pool for fusion. <strong>Finally,</strong> the dynamic matching-based <strong>H</strong>igh-level <strong>F</strong>eature <strong>R</strong>ecombination (HFR) is designed to tackle the third challenge, which uses the linguistic feature as an anchor to achieve semantically controllable explicit alignment and flexible implicit alignment by matching the most relevant features. Experimental results show our model achieves state-of-the-art performance on CMU-MOSI and CMU-MOSEI datasets, and demonstrates generalization capability on CH-SIMS dataset.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129805"},"PeriodicalIF":7.5,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159221","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":"Corrigendum to “Dual-branch hierarchical feature fusion network for video source camera identification” [Expert Systems with Applications 297 (Part B) (2026) 129475]","authors":"Bo Wang ,&nbsp;Jiaqi Chi ,&nbsp;Zhuocheng Wu ,&nbsp;Huimin Liu ,&nbsp;Wei Wang","doi":"10.1016/j.eswa.2025.129777","DOIUrl":"10.1016/j.eswa.2025.129777","url":null,"abstract":"","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"296 ","pages":"Article 129777"},"PeriodicalIF":7.5,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145218963","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":"MSIDiff:Multi-stage interaction-aware diffusion model for protein-specific 3D molecule generation","authors":"Yaoxiang Zhang,&nbsp;Junteng Ma,&nbsp;Ze Zhang,&nbsp;Zhaoyang Dong,&nbsp;Shuang Wang","doi":"10.1016/j.eswa.2025.129820","DOIUrl":"10.1016/j.eswa.2025.129820","url":null,"abstract":"<div><div>Structure-based drug design (SBDD) focuses on developing 3D ligand molecules that bind with high affinity to specific protein targets, which requires the accurate capture of the complex interactions between proteins and ligands. Although existing diffusion models have demonstrated potential in molecular generation tasks, they typically consider only a single stage of the generation process. This limitation prevents them from integrating the multi-stage protein-ligand interaction information from both forward and reverse processes, which may negatively impact the binding affinity of the generated molecules. To address this problem, MSIDiff (<strong>M</strong>ulti-<strong>S</strong>tage <strong>I</strong>nteraction-Aware <strong>Diff</strong>usion Model), a multi-stage interaction-aware diffusion model for protein-specific molecule generation, is proposed. MSIDiff leverages the pre-trained model MSINet to extract authentic protein-ligand interaction information during the initial diffusion stage and incorporates this information into the reverse process to ensure that the generated molecules accurately interact with target proteins. Through a scoring mechanism, MSIDiff filters key nodes to extract crucial protein-ligand interaction data and employs a GRU-based cross-layer interaction update module to recursively integrate information across different denoising stages, facilitating effective cross-layer information transmission. Experimental results on the CrossDocked2020 dataset show that MSIDiff can generate molecules with more realistic 3D structures and higher binding affinity to protein targets, achieving an Avg. Vina Score of up to -6.36, while maintaining appropriate molecular properties.Our code and data are available at: <span><span>https://github.com/zhangyaoxiang/MSIDiff</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129820"},"PeriodicalIF":7.5,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222003","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 efficient deep template matching and in-plane pose estimation method via template-aware dynamic convolution","authors":"Ke Jia ,&nbsp;Ji Zhou ,&nbsp;Hanxin Li ,&nbsp;Zhigan Zhou ,&nbsp;Haojie Chu ,&nbsp;Xiaojie Li","doi":"10.1016/j.eswa.2025.129813","DOIUrl":"10.1016/j.eswa.2025.129813","url":null,"abstract":"<div><div>In industrial inspection and component alignment tasks, template matching requires efficient estimation of a target’s position and geometric state (rotation and scaling) under complex backgrounds to support precise downstream operations. Traditional methods rely on exhaustive enumeration of angles and scales, leading to low efficiency under compound transformations. Meanwhile, most deep learning-based approaches only estimate similarity scores without explicitly modeling geometric pose, making them inadequate for real-world deployment. To overcome these limitations, we propose a lightweight end-to-end framework that reformulates template matching as joint localization and geometric regression, outputting the center coordinates, rotation angle, and independent horizontal and vertical scales. A Template-Aware Dynamic Convolution Module (TDCM) dynamically injects template features at inference to guide generalizable matching. The compact network integrates depthwise separable convolutions and pixel shuffle for efficient matching. To enable geometric-annotation-free training, we introduce a rotation-shear-based augmentation strategy with structure-aware pseudo labels. A lightweight refinement module further improves angle and scale precision via local optimization. Experiments show our 3.07M model achieves high precision and <span><math><mo>∼</mo></math></span>14 ms inference under compound transformations. It also demonstrates strong robustness in small-template and multi-object scenarios, making it highly suitable for deployment in real-time industrial applications. The code is available at: <span><span>https://github.com/ZhouJ6610/PoseMatch-TDCM</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129813"},"PeriodicalIF":7.5,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145189984","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-temporal ensemble for few-shot action recognition","authors":"Zhen Jiang,&nbsp;Jianlong Sun,&nbsp;Haodong Liu,&nbsp;Haizhen Guan","doi":"10.1016/j.eswa.2025.129821","DOIUrl":"10.1016/j.eswa.2025.129821","url":null,"abstract":"<div><div>Few-Shot Action Recognition (FSAR) aims to recognize novel action classes with only a few labeled samples. Due to the scarcity of labeled data, FSAR models suffer from high variance and low confidence. To address this issue, this paper first introduces ensemble learning into the field of FSAR, leveraging the diversity among multiple temporal action representations to generate base models. Specifically, we propose a Multi-Temporal Ensemble (MTE) method for FSAR. By combining sub-sequences of video frames of various lengths (i.e., tuples), MTE creates multiple sets of action representations and generates base models based on these representations. All base models share a single embedding network to learn frame-level features. The proposed method adaptively captures temporal relations with different lengths and speeds while avoiding the computational cost of training multiple deep neural networks. Furthermore, we introduce a Short-term Temporal Modeling Module (STMM) that uses self-attention to highlight frames with high variation, enhancing short-term temporal representation at the frame level. The proposed method has been validated on four benchmark datasets. Extensive experimental results demonstrate that MTE outperforms 26 state-of-the-art FSAR methods. The source code is available at <span><span>https://github.com/CharmainCahill/MTE.git</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129821"},"PeriodicalIF":7.5,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222093","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-scale fusion graph convolutional networks","authors":"Zhi Kong,&nbsp;Jie Ren,&nbsp;Lifu Wang,&nbsp;Ge Guo","doi":"10.1016/j.eswa.2025.129815","DOIUrl":"10.1016/j.eswa.2025.129815","url":null,"abstract":"<div><div>Graph analysis methods, as important tools for mining complex information, have made remarkable progress driven by graph neural networks (GNNs). However, existing approaches still face challenges in handling complex topological structures and multi-dimensional node features, making it difficult to fully capture deep-level feature and structural information. When analyzing attribute networks, a key challenge is how to effectively integrate node attribute features with graph topological structure information. To address this issue, this paper proposes a multi-scale fusion graph convolutional network (MSF-GCN) method. This method combines shallow and deep convolution strategies while adaptively fusing information across three parallel channels — the original topological structure, a feature-derived graph, and a deep-combination channel that captures shared depth information between them. An autoencoder is employed to reconstruct the adjacency matrix, enhancing the representation capability of the network. Additionally, an attention mechanism is introduced to dynamically assign weights to attribute and structural features at different scales, optimizing node representation. Experimental results demonstrate that, in node classification tasks across multiple benchmark datasets, MSF-GCN achieves outstanding performance, strongly validating the effectiveness and robustness of the proposed method.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"298 ","pages":"Article 129815"},"PeriodicalIF":7.5,"publicationDate":"2025-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221879","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}