Ting Zhang , Lihua Zhou , Xinchao Lu , Pei Zhang , Lizhen Wang
{"title":"HinMAD3R: Representation learning on heterogeneous information networks via multiple attentions with dual dropout and dual residual","authors":"Ting Zhang , Lihua Zhou , Xinchao Lu , Pei Zhang , Lizhen Wang","doi":"10.1016/j.eswa.2025.127674","DOIUrl":"10.1016/j.eswa.2025.127674","url":null,"abstract":"<div><div>Heterogeneous information networks (HINs) contain rich semantic information, and effectively utilizing this information can enhance the quality of representation learning. Existing models based on the message-passing paradigm typically focus on either node-type or edge-type information, neglecting the synergistic effect of various heterogeneous information. Moreover, these models are prone to over-smoothing as network depth increases, which degrades both performance and generalization ability. To address these issues, we propose a novel multiple attention mechanism that simultaneously considers node-features, node-types, and edge-types, aiming to maximize the utilization of diverse semantic information. Additionally, we introduce dual dropout and dual residual strategies to mitigate the over-smoothing problem and enhance the model’s generalization capability. Extensive experiments conducted on seven datasets demonstrate that the proposed model outperforms state-of-the-art baselines.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127674"},"PeriodicalIF":7.5,"publicationDate":"2025-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143869637","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}
Madjid Tavana , Mehdi Toloo , Francisco J. Santos-Arteaga , Hajar Farnoudkia , Violeta Cvetkoska
{"title":"Time envelopment analysis: A new method for effectively incorporating time series in data envelopment analysis","authors":"Madjid Tavana , Mehdi Toloo , Francisco J. Santos-Arteaga , Hajar Farnoudkia , Violeta Cvetkoska","doi":"10.1016/j.eswa.2025.127791","DOIUrl":"10.1016/j.eswa.2025.127791","url":null,"abstract":"<div><div>Data envelopment analysis (DEA) is a non-parametric tool for empirically evaluating the relative efficiency of homogeneous organizational units, i.e., decision-making units, by estimating the production frontiers. Time series analysis is a statistical technique that considers a series of data collected chronologically over time intervals. This study introduces a three-stage method, Time Envelopment Analysis (TEA), to effectively integrate time series analysis into DEA. The three-stage method includes a first-order autoregressive (AR(1)) model followed by DEA and ordinary least squares (OLS). The performance of the TEA method with four different values for the AR(1) parameters is compared with the DEA-OLS procedure using extensive Monte Carlo simulations. The simulation results show that the TEA method outperforms the DEA-OLS procedure. We further demonstrate that TEA is more accurate when the autoregressive parameter is smaller, particularly in scenarios defined by a progressive decrease in the impact of technical inefficiencies. We evaluate the proposed TEA method using a real-world healthcare dataset from 63 countries by estimating the effect of contextual variables on each country’s productivity.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127791"},"PeriodicalIF":7.5,"publicationDate":"2025-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143876888","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}
Caiwei Liu , Libin Tian , Pengfei Wang , Qian-Qian Yu , Li Song , Jijun Miao
{"title":"Non-destructive detection and quantification of corrosion damage in coated steel components with different illumination conditions","authors":"Caiwei Liu , Libin Tian , Pengfei Wang , Qian-Qian Yu , Li Song , Jijun Miao","doi":"10.1016/j.eswa.2025.127854","DOIUrl":"10.1016/j.eswa.2025.127854","url":null,"abstract":"<div><div>Existing deep learning-based detection methods for corrosion damage in steel structures are mostly applicable under normal lighting conditions and lack an association between detection results and damage levels. Focuses on coated corrosion steel components under various illumination conditions, this paper presents a YOLOv8s-G network tailored for pixel-level image segmentation and quantification of corrosion damage. A dataset of 1299 images of corroded steel components with different illumination conditions was captured in a field steel structure workshop. Furthermore, the ability of network to extract multi-scale corrosion features across various illumination conditions was enhanced by integrating the C2f-S module and fusion splicing method. The advancement and generalization of YOLOv8s-G were verified through comparisons with other state-of-the-art networks and tests on public datasets. Finally, the ratio of the corrosion area to the cross-sectional area of the steel component was calculated using morphological image operations, quantifying the relative area occupied by corrosion. The accuracy of this quantification method was further validated through comparison with filed measurements. Our research can enhance the reliability of decision-making regarding steel structural corrosion damage.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127854"},"PeriodicalIF":7.5,"publicationDate":"2025-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143874002","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":"Monocular visual semantic understanding system for real-time internal damage detection","authors":"Bian Xu , Tian Biwan , Yu Yangyang","doi":"10.1016/j.eswa.2025.127734","DOIUrl":"10.1016/j.eswa.2025.127734","url":null,"abstract":"<div><div>To quickly and non-destructively obtain internal damage data of components, endoscope technology based on machine vision has been widely utilized. However, it is still challenging to measure the internal damage of equipment intelligently and accurately in real time, especially for the internal damage of precision equipment. Therefore, an intelligent endoscope detection system based on monocular visual semantic understanding is proposed in this paper. In this system, a self-developed microprobe structure is used to construct a multi-frame full convolutional network model through multi-scale feature coupling mechanism, which effectively overcomes the feature degradation caused by low contrast imaging and uneven illumination during internal detection. As a result, it enables the automatic identification of the target region and the high − precision measurement of regional geometric parameters. Experimental results demonstrate that the average absolute error of damage size measurement is 0.029 mm, with a standard deviation of 0.0236. The average <em>mIoU</em> is at least 3.3 % higher than other detection methods covered in this article, and the accuracy of damage measurement is improved by about 10 %. It can realize automatic and intelligent defect identification and measurement, and meet the requirements of real-time measurement on site.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127734"},"PeriodicalIF":7.5,"publicationDate":"2025-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143864690","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}
Feng Wang , Shicheng Li , Shanshui Niu , Haoran Yang , Xiaodong Li , Xiaotie Deng
{"title":"A Survey on recent advances in reinforcement learning for intelligent investment decision-making optimization","authors":"Feng Wang , Shicheng Li , Shanshui Niu , Haoran Yang , Xiaodong Li , Xiaotie Deng","doi":"10.1016/j.eswa.2025.127540","DOIUrl":"10.1016/j.eswa.2025.127540","url":null,"abstract":"<div><div>Reinforcement learning (RL) has emerged as a powerful tool for optimizing intelligent investment decision-making. With the rapid evolution of financial markets, traditional approaches often struggle to effectively analyze the vast and complex datasets involved. RL-based methods address these challenges by leveraging neural networks to process large-scale financial data, dynamically interacting with market environments to refine strategies, and designing tailored reward functions to achieve diverse investment objectives. This paper provides a comprehensive review of recent advancements in RL for investment decision-making, with a focus on four key areas, i.e., portfolio selection, trade execution, options hedging, and market making. These four problems represent highly challenging instances of multi-stage , multi-objective decision optimization in investment, highlighting the strengths of RL-based methods in effectively balancing trade-offs among different objectives over time. Detailed comparison work of state-of-the-art RL-based methods is presented, analyzing the action spaces, state representations, reward structures, and neural network architectures. Finally, the paper discusses some new challenges and point out some directions for future research in the field.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127540"},"PeriodicalIF":7.5,"publicationDate":"2025-04-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143859168","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":"TRGNet: a deep transfer learning approach for software defect prediction","authors":"Meetesh Nevendra , Pradeep Singh","doi":"10.1016/j.eswa.2025.127799","DOIUrl":"10.1016/j.eswa.2025.127799","url":null,"abstract":"<div><div>Software defect prediction (SDP) aims to automatically locate defective modules to find bugs and prioritise testing efforts. Researchers are now shifting into semantic features in order to develop predictive models for accurate prediction by using deep learning. But the source code conversion into the semantic feature fails to capture the essential features and correlation. This often degrades the performance of the prediction model. However, well-known authors have already shown the importance of software module metrics for software defect prediction. To take the advantage of software metrics via deep transfer learning in this paper, software module metrics are transformed into images. We proposed the TRGNet model, which extracts transferable features from source projects using pre-trained GoogLeNet and consolidates with a <em>meta</em>-estimator to minimize the divergence in sample distributions between projects. In this model, we feed the transformed image file of software modules to train it for within-project defect prediction (WPDP) and cross-project defect prediction (CPDP). The experimental results with AlexNet, ResNet, SqueezeNet, and other state-of-the-art models indicate that the proposed TRGNet model significantly improves the state-of-the-art defect prediction task by 13.31 % in WPDP and 16.88 % in CPDP scenarios. Moreover, the computational cost analysis reveals that TRGNet significantly reduces memory utilization while maintaining competitive training and inference times compared to other deep learning models, making it a highly efficient and scalable approach for SDP.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127799"},"PeriodicalIF":7.5,"publicationDate":"2025-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143869513","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":"Sudoku puzzle generation using mathematical programming and heuristics: Puzzle construction and game development","authors":"Tugce Ates, Fatih Cavdur","doi":"10.1016/j.eswa.2025.127710","DOIUrl":"10.1016/j.eswa.2025.127710","url":null,"abstract":"<div><div>In this study, first, a new mathematical programming formulation for generating Sudoku puzzles is proposed. It is possible to generate specially-configured puzzle instances using the proposed formulation which is flexible enough to control not only the numbers of the Sudoku matrix entries shown in each column, row and sub-matrix, but also the times each number appears by setting up the corresponding model parameters accordingly. The initially developed non-linear program with a quadratic constraint is reformulated as a linear-integer program by using appropriate variate transformations. The resulting mathematical program is then solved to generate Sudoku puzzles and its computational performance is analyzed through computational experiments. It is noted that the formulation is fast enough to generate Sudoku puzzles in reasonable time periods using a commercial solver on a personal computer. The study then discusses how to ensure the uniqueness of a solution for a puzzle instance generated by a hybrid approach that integrates the mathematical program with a heuristic algorithm. In the final part of the study, the idea of the proposed hybrid approach is extended and a backtracking algorithm-based puzzle generation procedure is designed and implemented by developing a standalone mobile-web game application.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127710"},"PeriodicalIF":7.5,"publicationDate":"2025-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143874003","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-class financial distress prediction based on hybrid feature selection and improved stacking ensemble model","authors":"Xiaofang Chen , Jiaming Liu , Chong Wu","doi":"10.1016/j.eswa.2025.127832","DOIUrl":"10.1016/j.eswa.2025.127832","url":null,"abstract":"<div><div>Multi-class financial distress prediction (FDP) can accurately assess the corporate financial status. Improving its prediction performance is the academic focus. Feature selection and classifier models play a crucial role in the multi-class FDP model. Therefore, this paper proposes a new hybrid feature selection and an improved stacking ensemble model. The hybrid feature selection uses information gain and an improved particle swarm optimization to filter the indicators. The hyperopt hyperparameter optimization method is used to optimize the base learners of stacking ensemble model; The F1-score weighted optimization method is designed for dealing with the discrepancies of the base learners; To objectively solve the combination configuration problem of stacking ensemble model, a constrained genetic algorithm is proposed. The Chinese listed companies are used as research objects for empirical research. The results show that the hybrid feature selection outperforms other feature selection. The F1-score weighted optimized model has 8.97% higher accuracy than the unweighted optimized model. The proposed model performs better in terms of accuracy, robustness, and sensitivity compared to the baseline models and the classifier models in existing multi-class FDP studies. The proposed hybrid feature selection and the improved stacking ensemble model provide new and reliable research ideas for multi-class FDP.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127832"},"PeriodicalIF":7.5,"publicationDate":"2025-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143874005","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}
Yan He , Weiwei Zhang , Hai Zhang , Jinde Cao , Mahmoud Abdel-Aty
{"title":"Projective synchronization results of fractional order quaternion valued neural networks with proportional delay under event-triggered control","authors":"Yan He , Weiwei Zhang , Hai Zhang , Jinde Cao , Mahmoud Abdel-Aty","doi":"10.1016/j.eswa.2025.127643","DOIUrl":"10.1016/j.eswa.2025.127643","url":null,"abstract":"<div><div>This paper explores the projective synchronization (PS) of fractional order neural networks (FONNs) with proportional delay in quaternion through non-decomposition method. A controller with proportional delay is designed from the perspective of reducing the times of controller updates and computational costs, and corresponding event-triggered conditions are provided. The criteria for achieving PS of the systems are obtained through techniques such as mean value inequality and Razumikhin’s theorem. On the basis of this derivation, the conditions for systems to achieve PS are explored when two consecutive releases are fixed. In addition, the positive lower bound of the inter-event time is derived, Zeno behavior can be excluded. Finally, the validity of the theoretical results is obtained through simulation. Additionally, the application of the system studied in this paper in image encryption and decryption is presented.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127643"},"PeriodicalIF":7.5,"publicationDate":"2025-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143864798","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}
Yijie Wang , Runqing Wang , Jian Sun , Fang Deng , Gang Wang , Jie Chen
{"title":"Attention enhanced reinforcement learning for flexible job shop scheduling with transportation constraints","authors":"Yijie Wang , Runqing Wang , Jian Sun , Fang Deng , Gang Wang , Jie Chen","doi":"10.1016/j.eswa.2025.127671","DOIUrl":"10.1016/j.eswa.2025.127671","url":null,"abstract":"<div><div>In smart manufacturing systems, the flexible job-shop scheduling problem with transportation constraints (FJSPT) is a critical challenge that can significantly improve production efficiency. FJSPT extends the traditional flexible job-shop scheduling problem (FJSP) by integrating the scheduling of automated guided vehicles (AGVs) to transport intermediate products between machines. Recent advances in data-driven methods, particularly deep reinforcement learning (DRL), have addressed challenging combinatorial optimization problems. DRL effectively solves discrete optimization problems by generating high-quality solutions within reasonable time. This paper presents an end-to-end DRL approach for the simultaneous scheduling of machines and AGVs in FJSPT. To apply DRL to the FJSPT, this paper first formulates a Markov decision process (MDP) model. The action space combines operation selection, machine assignment, and AGV planning. To capture problem characteristics, the scheduling agent uses a graph attention network (GAT) and multi-layer perceptron (MLP) for feature extraction, combined with proximal policy optimization (PPO) for stable training. Experimental evaluations conducted on synthetic data and public instances demonstrate that the proposed method outperforms dispatching rules and state-of-the-art models in both scheduling performance and computational efficiency.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"282 ","pages":"Article 127671"},"PeriodicalIF":7.5,"publicationDate":"2025-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143864802","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}