Applied Soft Computing最新文献

{"title":"Online ensemble learning-based anomaly detection for IoT systems","authors":"Yafeng Wu,&nbsp;Lan Liu,&nbsp;Yongjie Yu,&nbsp;Guiming Chen,&nbsp;Junhan Hu","doi":"10.1016/j.asoc.2025.112931","DOIUrl":"10.1016/j.asoc.2025.112931","url":null,"abstract":"<div><div>In the modern era of digital transformation, the evolution of fifth-generation (5G) wireless networks has played a pivotal role in revolutionizing communication technology and accelerating the growth of smart technology applications. As an integral element of smart technology, the Internet of Things (IoT) grapples with the problem of limited hardware performance. Cloud and fog computing-based IoT systems offer an effective solution but often encounter concept drift issues in real-time data processing due to the dynamic and imbalanced nature of IoT environments, leading to performance degradation. In this study, we propose a novel framework for drift-adaptive ensemble learning called the Adaptive Exponentially Weighted Average Ensemble (AEWAE), which consists of three stages: IoT data preprocessing, base model learning, and online ensembling. It integrates four advanced online learning methods within an ensemble approach. The crucial parameter of the AEWAE method is fine-tuned using the Particle Swarm Optimization (PSO) technique. Experimental results on four public datasets demonstrate that AEWAE-based anomaly detection effectively detects concept drift and identifies anomalies in imbalanced IoT data streams, outperforming other baseline methods in terms of accuracy, F1 score, false alarm rate (FAR), and latency.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"173 ","pages":"Article 112931"},"PeriodicalIF":7.2,"publicationDate":"2025-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143550941","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-objective evolutionary algorithm with two balancing mechanisms for heterogeneous UAV swarm path planning","authors":"Xiuju Xu ,&nbsp;Chengyu Xie ,&nbsp;Linru Ma ,&nbsp;Lin Yang ,&nbsp;Tao Zhang","doi":"10.1016/j.asoc.2025.112927","DOIUrl":"10.1016/j.asoc.2025.112927","url":null,"abstract":"<div><div>Unmanned aerial vehicle (UAV) swarm path planning involves creating efficient routes based on task requirements to enable collaborative flight. Compared to homogeneous UAV swarm, the application scenarios of heterogeneous UAV swarm have become increasingly widespread. They can fully leverage the various capabilities of drones and show higher economic benefits. Existing research mainly focuses on homogeneous UAV swarms, and the model for uniformly describing heterogeneous UAV swarm from a functional perspective is insufficient. Differences in dynamic constraints and energy consumption models create challenges for accurately characterizing the path planning problem of heterogeneous UAV swarm. To supplement the above deficiencies, this article designs the scenario and composition structure of heterogeneous UAV swarm. The path-planning problem of heterogeneous UAV swarm is modeled as a multi-objective optimization (MOO) problem, in which a comprehensive energy consumption objective is constructed. To better balance multiple objectives and obtain high-quality solutions, a MOO evolutionary algorithm based on heterogeneous UAV swarm, namely HMOEA, is proposed. Specifically, HMOEA is implemented by combining the proposed two strategies. To verify the model’s feasibility and the algorithm’s effectiveness, numerical simulations and prototype simulations are provided. In numerical simulations, the proposed algorithm was compared with various advanced algorithms, i.e., NSGA-II, CIACO, AP-GWO, CL-DMSPSO, and DSNSGA-III, in two designed terrain problems. The results demonstrate that HMOEA not only outperforms the compared algorithms on convergence and diversity indicators increased over 4% and 2% respectively. Normal flight results were achieved in the two scenarios served by the prototype simulation, namely, urban buildings and forest scenes. Specific implementation and application can be achieved in military or civilian scenarios like reconnaissance and strike missions, search and rescue missions. The proposed model can adapt to more task scenarios, and the proposed method can provide faster and higher quality results for heterogeneous UAV swarm routes. In actual deployment, adjusting model parameters and optimizing the computing environment according to application requirements are worth further investigation to achieve optimal effect.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"173 ","pages":"Article 112927"},"PeriodicalIF":7.2,"publicationDate":"2025-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143550943","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":"Dual residual learning of frequency fingerprints in detecting synthesized biomedical imagery","authors":"Misaj Sharafudeen,&nbsp;Vinod Chandra S.S.","doi":"10.1016/j.asoc.2025.112930","DOIUrl":"10.1016/j.asoc.2025.112930","url":null,"abstract":"<div><div>Artificial synthesis of biomedical imagery is an evolving threat yet under-addressed. The integrity of medical imaging is important for accurate diagnosis and treatment. This study addresses the potential threat of fabricated biomedical imagery, focusing on synthetic dermatological lesions and CT nodules. The Representation Similarity Matrix measured the quantitative authenticity to account for similarities of synthesized data with authentic data. The study explores traces of manipulation from frequency signatures of synthesized imagery. We propose a novel combinatorial architecture, the Dual Residual Network (DRN), capturing hidden residual traces from low-frequency fingerprints of synthetic data and exposing hidden forgeries. DRN achieves near-perfect detection rates with an accuracy of 98.80% for CT nodules and 98.97% for lesions. Equal Error Rates of the model on the two datasets exhibited a marginal improvement of 57.87% in the CT nodules compared to the skin lesions. Sensitivity and specificity play a significant role in medical diagnostics. The model achieved sensitivities of 99.31% and 98.45% and specificity of 98.80% and 99.60% for each dataset, respectively. Further verification of the frequency traces was performed by analyzing gradients in the target concepts that led to decision-making. This study equips the medical field with a powerful tool to combat the evolving threat of synthetic fraud, safeguarding patient and client safety. The potential of the technique extends beyond healthcare, offering a blueprint for tackling synthetic data across diverse domains.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"173 ","pages":"Article 112930"},"PeriodicalIF":7.2,"publicationDate":"2025-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143551048","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":"Automatic optimization for generating adversarial malware based on prioritized evolutionary computing","authors":"Yaochang Xu,&nbsp;Yong Fang,&nbsp;Yijia Xu,&nbsp;Zhan Wang","doi":"10.1016/j.asoc.2025.112933","DOIUrl":"10.1016/j.asoc.2025.112933","url":null,"abstract":"<div><div>Machine learning has been widely applied to malware detection tasks; but unfortunately, they exhibit significant vulnerability to adversarial attacks and can be easily circumvented using perturbation carefully crafted. Concurrently, we are witnessing a corresponding increase in the attention dedicated to adversarial attacks against malware detection models. Nevertheless, current research on adversarial examples still faces obstacles such as poor escape effectiveness and difficulty in preserving functionality. Particularly, greedily recruiting the best manipulations from a vast search space often leads to poor diversity of adversarial perturbation sequence. To rectify these shortcomings, this paper proposes an automated, continuously optimized approach for generating malware adversarial examples based on evolutionary computing. Our method filters effective action sequences from a large pool of random manipulations, assigning different priorities to different actions. The generation and optimization of adversarial examples are formalized as a sparse minimization optimization problem based on a fixed-length action vector. We introduce AOP-Mal, a novel genetic framework to automatically generate and optimize adversarial examples. The initialization and evolution of the population depend on the priority of actions, as well as the proposed novel evolutionary operator. The experimental results demonstrate that our attack strategy effectively bypasses the detection mechanisms and outperforms most state-of-the-art malware adversarial frameworks. Our hope is to help researchers understand the intentions of attackers and explore more powerful defense mechanisms.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"173 ","pages":"Article 112933"},"PeriodicalIF":7.2,"publicationDate":"2025-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143550940","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 random searching algorithm for efficiently solving the connectivity-oriented robust optimization problem on large-scale networked systems","authors":"Wei Wei,&nbsp;Guobin Sun,&nbsp;Qinghui Zhang","doi":"10.1016/j.asoc.2025.112924","DOIUrl":"10.1016/j.asoc.2025.112924","url":null,"abstract":"<div><div>By consolidating part of the links to be invulnerable, there will be no connectivity degradation in a network under expected network failure intensity. Although existing link consolidation methods can handle large-scale networks, their solutions are far from optimal. Redundancy in existing solutions can be quantified by the connectivity of the pure graph consisting of the necessary subset of links, and existing methods improve pure graph connectivity to far above the expected value. Fortunately, we have found a special superset of link cuts, and proved that it can reduce consolidation links by removing the right set of links from existing solutions while maintaining the desired connectivity. In response to the high complexity of searching for the optimal superset, we found one kind of superset that is close to the optimal solution and easy to locate, significantly reducing the number of links that need to be consolidated with a slight increase in preprocessing overhead. Experiments have shown that in large networks, the algorithm can provide a protection effect of over 99.9%, and can lead to 60% overhead savings compared to existing high-speed algorithms under the same computing time. On small-scale networks where the optimal algorithm is feasible, the average additional cost compared to the optimal result can be controlled within 1%. Thus, while ensuring accuracy, it can further approach the optimal solution compared to existing algorithms, significantly reducing the overhead of infrastructure consolidation.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"173 ","pages":"Article 112924"},"PeriodicalIF":7.2,"publicationDate":"2025-03-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143550939","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 unified transductive and inductive learning framework for Few-Shot Learning using Graph Neural Networks","authors":"Jie Chang ,&nbsp;Haodong Ren ,&nbsp;Zuoyong Li ,&nbsp;Yinlong Xu ,&nbsp;Taotao Lai","doi":"10.1016/j.asoc.2025.112928","DOIUrl":"10.1016/j.asoc.2025.112928","url":null,"abstract":"<div><div>Graph Neural Networks (GNNs) have shown their effectiveness in integrating feature embeddings for image and video processing tasks. While initially developed for inductive learning, GNNs have been extended to support transductive learning, enabling them to learn from partially labeled graphs. However, the combination of transductive and inductive learning in existing GNN-based models lacks proper theoretical justification, and GNNs with information propagation mechanisms often encounter the over-smoothing problem, especially in Few-Shot Learning (FSL) tasks. In this paper, we propose a unified transductive and inductive learning GNN model named FGCN for FSL tasks. The proposed FGCN differentiates between the roles of inductive and transductive learning, while quantifying the contributions of intra-properties within entities and inner-relationships between neighboring entities. By addressing the over-smoothing problem comprehensively, the FGCN offers a promising approach for FSL tasks. Our findings demonstrate that the proposed FGCN model achieves a significant improvement in accuracy over state-of-the-art methods, as evidenced by experiments on four standard Few-Shot Learning benchmarks. For example, in the 5-Way 5-Shot scenario, the proposed FGCN achieved an accuracy increase of 7.70% on the Mini-ImageNet, compared to the state-of-the-art result obtained by the MCGN model.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"173 ","pages":"Article 112928"},"PeriodicalIF":7.2,"publicationDate":"2025-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143551047","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":"Heterogeneous collaborative filtering contrastive learning for social recommendation","authors":"Chaojun Meng ,&nbsp;Changfan Pan ,&nbsp;Hongji Shu ,&nbsp;Qing Wang ,&nbsp;Hanghui Guo ,&nbsp;Jia Zhu","doi":"10.1016/j.asoc.2025.112934","DOIUrl":"10.1016/j.asoc.2025.112934","url":null,"abstract":"<div><div>Collaborative filtering methods based on Graph Neural Networks (GNNs) have gained increasing popularity in recommendation systems. These methods enhance the representation of users and items by leveraging the information of graph structure from interaction data, improving recommendation performance. However, they often face limitations due to the data sparsity issue that is common in recommendation systems. In the constructed user–item heterogeneous bipartite graph, sparse interaction data leads to a scarcity of neighbor nodes impeding the acquisition of sufficient collaborative signals via the message-passing mechanism among these neighbor nodes. We have observed that users and items can be grouped according to characteristic similarities. These groups’ common feature information can serve as supplementary data to aid in the embedding learning. So, we present the Heterogeneous Collaborative Filtering Contrastive Learning (HCFCL) method, which aims to extract two types of heterogeneous collaborative signals from interaction data: those based on neighbor nodes and those based on group features. Specifically, we design an embedding generative hypergraph network to extract group common feature information founded on the heterogeneous bipartite graph. The group common feature information is transferred via a meta network and personalized bridge functions according to individual characteristics. Additionally, the HCFCL model, combined with contrastive learning, captures the consistency of the heterogeneous collaborative signals to enhance representation. The experiment demonstrates the superior performance of the HCFCL model compared to other methods evaluated on three public datasets, demonstrating excellent and stable performance in mitigating the data sparsity issue.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"173 ","pages":"Article 112934"},"PeriodicalIF":7.2,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143551049","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":"Fill-UNet: Extended composite semantic segmentation","authors":"Qunpo Liu ,&nbsp;Yi Zhao ,&nbsp;Weiping Ding ,&nbsp;Xuhui Bu ,&nbsp;Naohiko Hanajima","doi":"10.1016/j.asoc.2025.112891","DOIUrl":"10.1016/j.asoc.2025.112891","url":null,"abstract":"<div><div>The accuracy of image segmentation directly affects the precision of object recognition. To address the limitations of the U-Net model in capturing global contextual information and leveraging deep semantic features, an extended composite semantic segmentation model, Fill-UNet, is proposed for extracting deep semantic features. The proposed Semantic Collaborative Filtering Attention Module (SCFAM) enhances the model's ability to perceive channel information by compressing channel-direction features while preserving high-resolution mask semantic details. The integration of a Transformer structure into the encoder-decoder connections enables the extraction of deep pixel-level features. Furthermore, the designed Multi-Scale Semantic Feature Association mechanism (MSFA), which combines short and long skip connections with a Parallel Weighted Fusion Module (PWFM), strengthens multi-scale feature fusion, particularly for small objects. Extensive experiments conducted on the PASCAL VOC2012, Cityscapes, CamVid, and Pascal Context datasets demonstrate that Fill-UNet achieves significant improvements in pixel segmentation accuracy while offering a slight increase in segmentation speed compared to state-of-the-art methods. The code will be available at <span><span>https://github.com/ZzYH-i/Fill-UNet</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"172 ","pages":"Article 112891"},"PeriodicalIF":7.2,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143508508","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 interpretable Dahl-LRN neural-network for accurately modelling the systems with rate-dependent asymmetric hysteresis","authors":"Lei Ni ,&nbsp;Hongfei Wang ,&nbsp;Guoqiang Chen ,&nbsp;Lanqiang Zhang ,&nbsp;Na Yao ,&nbsp;Geng Wang","doi":"10.1016/j.asoc.2025.112936","DOIUrl":"10.1016/j.asoc.2025.112936","url":null,"abstract":"<div><div>The motion accuracy and stability of piezoelectric positioning systems are significantly compromised by inherent hysteresis and other nonlinearities. This paper presents an innovative method integrating the Dahl model with Layer Recurrent Neural Networks (LRN) to model piezoelectric actuators accurately. Initially, the Dahl model is reformulated into a neural network structure, resulting in the Dahl Neural Network (DahlNN), which strictly adheres to the underlying mathematical equations. The weights of this network directly correspond to the parameters of the Dahl equations, thereby creating a transparent neural network architecture with clear physical significance and interpretability. Subsequently, the DahlNN is enhanced by incorporating feedback mechanisms and recurrent effects from LRN, improving its ability to describe asymmetric and rate-dependent hysteresis characteristics. Extensive experiments demonstrate that, compared to LRN models without physical knowledge guidance, the proposed Dahl-LRN model reduces peak-to-valley fluctuations by 70 % and decreases the average error by approximately 97.3 %, with only a 5 % increase in computational time while maintaining interpretability and achieving superior modelling performance. Through this approach, this paper aims to provide a novel perspective on leveraging physical information to advance the application of deep learning in modelling complex physical phenomena.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"173 ","pages":"Article 112936"},"PeriodicalIF":7.2,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143551050","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":"Machine-learning component for multi-start metaheuristics to solve the capacitated vehicle routing problem","authors":"Juan Pablo Mesa ,&nbsp;Alejandro Montoya ,&nbsp;Raul Ramos-Pollan ,&nbsp;Mauricio Toro","doi":"10.1016/j.asoc.2025.112916","DOIUrl":"10.1016/j.asoc.2025.112916","url":null,"abstract":"<div><div>Multi-Start metaheuristics (MSM) are commonly used to solve vehicle routing problems (VRPs). These methods create different initial solutions and improve them through local-search. The goal of these methods is to deliver the best solution found. We introduce initial-solution classification (ISC) to predict if a local-search algorithm should be applied to initial solutions in MSM. This leads to a faster convergence of MSM and higher-quality solutions when computation time is limited. In this work, we extract features of a capacitated VRP (CVRP) solution, by transforming the structure of a solution into quantitative metrics (i.e.number of customers in each route, average compactness of a route, or number of intersections between routes). With these features and a machine-learning classifier (random forest), we show how ISC – significantly – improves the performance of greedy randomized adaptive search procedure (GRASP), over benchmark instances from the CVRP literature. With the objective of evaluating ISC’s performance with different local-search algorithms, we implemented a local-search composed of classical neighborhoods from the literature and another local-search with only a variation of Ruin-and-Recreate. In both cases, ISC significantly improves the quality of the solutions found in almost all the evaluated instances.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"173 ","pages":"Article 112916"},"PeriodicalIF":7.2,"publicationDate":"2025-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143550896","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}