Applied Soft Computing最新文献

{"title":"Projection-based comprehensive multi-view clustering with smooth regularization","authors":"Xiaoqian Zhang ,&nbsp;Jinghao Li ,&nbsp;Shuai Zhao ,&nbsp;Yilu Zheng ,&nbsp;Lei Pu ,&nbsp;Yufeng Chen","doi":"10.1016/j.asoc.2025.113025","DOIUrl":"10.1016/j.asoc.2025.113025","url":null,"abstract":"<div><div>By mining latent representations of data, subspace clustering methods can be more accurate and robust. However, such methods face the following limitations: they lack the ability to explicitly preserve the relationships between individuals within the original data cluster when constructing low dimensional latent representations. Specifically, during data dimensionality reduction, it cannot ensure that the retained information is relevant information between samples. Secondly, data dimensionality reduction may lead to the loss of key information within some samples, affecting the construction of self-representation matrices and reducing clustering performance. To solve these two problems, a new multi-view clustering method is proposed, Projection-based comprehensive multi-view clustering with smooth regularization (PCMCS). We design a smooth regular term for the projection matrix, which can make the data after dimensionality-reduced retain the grouping effect of the original data. Then, we capture the representation matrix of the original data and the data after dimensionality-reduced, and construct the resulting representation matrix as a tensor such that the two self-representation matrices are optimized with respect to each other, which to a certain extent neutralizes the interference of information loss, data redundancy, and noise in the construction of the representation matrices and improves the clustering performance. Experiments are conducted on 8 datasets, demonstrating the effectiveness of PCMCS.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113025"},"PeriodicalIF":7.2,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143776300","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":"Deep reinforcement learning for optimal firebreak placement in forest fire prevention","authors":"Lucas Murray ,&nbsp;Tatiana Castillo ,&nbsp;Isaac Martín de Diego ,&nbsp;Richard Weber ,&nbsp;José Ramón González-Olabarria ,&nbsp;Jordi García-Gonzalo ,&nbsp;Andrés Weintraub ,&nbsp;Jaime Carrasco-Barra","doi":"10.1016/j.asoc.2025.113043","DOIUrl":"10.1016/j.asoc.2025.113043","url":null,"abstract":"<div><div>The increasing frequency and intensity of large wildfires have become a significant natural hazard, requiring the development of advanced decision-support tools for resilient landscape design. Existing methods, such as Mixed Integer Programming and Stochastic Optimization, while effective, are computationally demanding. In this study, we introduce a novel Deep Reinforcement Learning (DRL) methodology to optimize the strategic placement of firebreaks across diverse landscapes. We employ Deep Q-Learning, Double Deep Q-Learning, and Dueling Double Deep Q-Learning, integrated with the Cell2Fire fire spread simulator and Convolutional Neural Networks. Our DRL agent successfully learns optimal firebreak locations, demonstrating superior performance compared to heuristics, especially after incorporating a pre-training loop. This improvement ranges between 1.59%–1.7% with respect to the heuristic, depending on the size of the instance, and 4.79%–6.81% when compared to a random solution. Our results highlight the potential of DRL for fire prevention, showing convergence with favorable results in cases as large as 40 × 40 cells. This study represents a pioneering application of reinforcement learning to fire prevention and landscape management.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113043"},"PeriodicalIF":7.2,"publicationDate":"2025-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143738479","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-label feature selection via label relaxation","authors":"Yuling Fan ,&nbsp;Peizhong Liu ,&nbsp;Jinghua Liu","doi":"10.1016/j.asoc.2025.113047","DOIUrl":"10.1016/j.asoc.2025.113047","url":null,"abstract":"<div><div>Multi-label feature selection (MFS) has emerged as a prevalent strategy to manage high-dimensional multi-label data. Most existing methods assume that the rigid binary label matrix can perfectly fit the pseudo-label matrix during the learning process, so as to preserve the structural information in raw data. However, the original label space with the limited freedom makes it challenging to accurately convert to the pseudo-label matrix. Additionally, most methods utilize different matrix to explore structural information, and ignore the connection of structural information. To tackle these problems, a novel method named multi-label feature selection via label relaxation (LRMFS) is proposed. LRMFS designs a label relaxation regression to transform the rigid binary label matrix into a slack variable matrix, allowing for a more flexible fitting relationship. By leveraging this flexible fitting, LRMFS decomposes the feature selection matrix to a structured subspace, which can learn the graph structures of both features and labels by graph Laplacian. These properties of LRMFS are converted to an objective function, and we further develop an alternative solution for the function optimization. Comparative experiments show that LRMFS exhibits superior performance than eight MFS methods on twelve multi-label data sets.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113047"},"PeriodicalIF":7.2,"publicationDate":"2025-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143738288","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":"Security modules of delegation methods in mobile cloud computing using probabilistic interval neutrosophic hesitant fuzzy set based decision-making model","authors":"Sendeyah Al Hantoobi ,&nbsp;A.A. Zaidan ,&nbsp;Hassan Abdulsattar Ibrahim ,&nbsp;Sarah Qahtan ,&nbsp;Muhammet Deveci ,&nbsp;Sinan Isik ,&nbsp;Hana Tomášková","doi":"10.1016/j.asoc.2025.113089","DOIUrl":"10.1016/j.asoc.2025.113089","url":null,"abstract":"<div><div>The security models of delegation methods in mobile cloud computing (MCC) play a crucial role in mitigating security threats when offloading device operations to the cloud while ensuring optimal performance. These threats include data breaches, unauthorized access, and loss of control over sensitive data. Although numerous security models have been developed for different delegation methods, categorized under layering, authentication, dynamic offloading, and encryption, none fully satisfy all development security criteria despite extensive research efforts. To address this gap, this paper proposes a decision-modeling approach to identify the most effective security modules for delegation methods in MCC. Such problem falls under multi-criteria decision-making (MCDM) due to (1) the presence of multiple development security criteria, (2) the varying importance of these criteria with inherent ambiguity and uncertainty, and (3) data variation. To achieve this, we developed a fuzzy-weighted zero-inconsistency method (FWZIC) under a probabilistic interval neutrosophic hesitant fuzzy set (PINHFS) and the evaluation based on distance from the average solution (EDAS) technique. In addition, the decision matrix in this paper is constructed by crossing 50 security modules from four delegation methods in MCC, categorized as follows: 23 encryption, 12 authentication, 9 layering, and 6 dynamic offloading modules, evaluated based on 13 development security criteria. The findings of PINHFS–FWZIC method indicate that <span><math><msub><mrow><mi>C</mi></mrow><mrow><mn>7</mn></mrow></msub></math></span> ‘Data Security’ was the most critical and sensitive evaluation criterion, while <span><math><msub><mrow><mi>C</mi></mrow><mrow><mn>1</mn></mrow></msub></math></span> ‘General Security Issue’ was the least sensitive. According to the EDAS results, SM_07 ranked highest for layering and encryption-based delegation methods. For dynamic offloading, SM_02 and SM_03 achieved the top ranking, while for authentication, SM_02 and SM_03 were also identified as the best security modules. To validate the robustness of the proposed approach, sensitivity analysis, Spearman’s correlation coefficient test, and comparative analysis were conducted. The results confirm that the proposed approach provides a comprehensive and reliable method for selecting optimal security modules in MCC delegation methods.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113089"},"PeriodicalIF":7.2,"publicationDate":"2025-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143776301","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Historical oracle bone character recognition through domain transfer and mutual learning","authors":"Zhengchen Li ,&nbsp;Xiuan Wan ,&nbsp;Jiahua Wu ,&nbsp;Shouyong Pan ,&nbsp;Yuchun Fang","doi":"10.1016/j.asoc.2025.113031","DOIUrl":"10.1016/j.asoc.2025.113031","url":null,"abstract":"<div><div>Oracle bone characters (OBCs) provide crucial primary evidence for studying the history of the Shang Dynasty and the evolution of Chinese characters. However, the difficulty of collecting and annotating authentic OBC images, the intra-class difference, and inter-class similarity pose significant challenges for OBC recognition. In this paper, we propose a domain transfer and mutual learning network (DTML) for cross-domain OBC recognition, which leverages OBCs handprinted and annotated by experts to recognize unlabeled authentic OBCs. The domain transfer strategy shifts authentic OBC samples to match the distribution of handprinted OBC samples and vice versa, preserving the domain-specific features of both handprinted and authentic OBCs while implicitly decreasing the distributional differences between the two domains. The mutual learning strategy leverages high-confidence samples from two distinct domain distributions to guide the training of each other’s domain models, facilitating the transfer of domain-specific knowledge between the two domains. Comprehensive testing shows that our approach establishes a new benchmark on the Oracle-241 dataset, surpassing the latest state-of-the-art recognition accuracy by 10.0%. Our code is temporarily available at <span><span>https://github.com/ycfang-lab/DTML</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113031"},"PeriodicalIF":7.2,"publicationDate":"2025-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143746219","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":"Greedy mechanism-based bi-objective optimization for green scheduling in manufacturing systems considering transportation","authors":"Zhu Wang ,&nbsp;Rongping Qiu ,&nbsp;Binghai Zhou","doi":"10.1016/j.asoc.2025.113093","DOIUrl":"10.1016/j.asoc.2025.113093","url":null,"abstract":"<div><div>This paper addresses scheduling challenges in hybrid flow manufacturing systems with crane transportation (HFMS-CT) driven by intelligent control, mass customization, and eco-friendly manufacturing. Unlike previous studies, it considers the interdependence between machine processing and crane transport, focusing on minimizing both makespan and energy consumption. A bi-objective mixed-integer programming model is developed, and the Epsilon-constraint method is used for small-scale cases. Given the NP-hardness, a modified multi-objective Harris Hawk optimization (MMOHHO) is proposed. It adopts greedy mechanisms by integrating Laplace crossover, tent-based chaotic mapping, elite selection, and nonlinear optimization strategy to balance exploration and exploitation capabilities. The proposed algorithm is compared with the Epsilon-constraint method and benchmark metaheuristics. The experimental results reveal that the proposed algorithm outperforms other methods regarding NPS, DPO, IGD, and ES evaluation metrics. Finally, an in-depth discussion is conducted using a real-world case study, offering valuable managerial insights and practical recommendations for implementation.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113093"},"PeriodicalIF":7.2,"publicationDate":"2025-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143746376","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":"Bug priority prediction using deep ensemble approach","authors":"P.G.S.M. Dharmakeerthi ,&nbsp;R.A.H.M. Rupasingha ,&nbsp;B.T.G.S. Kumara","doi":"10.1016/j.asoc.2025.113098","DOIUrl":"10.1016/j.asoc.2025.113098","url":null,"abstract":"<div><div>A software bug is a fault in the programming of software or an application. Bugs cause problems ranging from stability to operability and are typically the result of human error during the programming process. They could be the result of a mistake or error, as well as a fault or defect. Software bugs should be discovered during the testing stage of the software development life cycle, but some may go undetected until after deployment. When addressing a bug, it is critical to consider its priority, which is determined manually. However, it was a difficult task, and making the wrong decision could lead to major software failures. Therefore, the primary goal of this study is to propose an ensemble approach for predicting bug priority levels in bug reports. We make use of Bugzilla's dataset, which includes over 25,000 bug reports. After preprocessing the data, this study applies a variety of feature extraction techniques, including Glove, Word2Vec TF-IDF, and Doc2Vec. Then, a model that primarily employs seven architectures of Convolutional Neural Network (CNN) Algorithms, including AlexNet, LeNet, VGGNet, 1DCNN, ResNet, ZF Net, and DenseNet as the basic models. The five architectures with the highest accuracy were then used in the ensemble method, which included ResNet, DenseNet, LeNet, AlexNet, and 1DCNN, with the final results determined by the majority values. The ensemble approach performed with 79.18 % of the final accuracy result. Other architectures include AlexNet 77.1 %, ZF Net 44.50 %, VGG Net 39.30 %, 1DCNN 75.44 %, ResNet 77.34 %, DenseNet 77.32 %, and LeNet 48.58 %. It was discovered that the proposed ensemble model outperformed each algorithm. Finally, when a new bug is discovered, it can be added to the proposed model, which will then determine its priority level.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113098"},"PeriodicalIF":7.2,"publicationDate":"2025-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143735024","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-adaptive framework of reducing domain bias under distribution shift for semi-supervised domain generalization","authors":"Liangqing Hu ,&nbsp;Zuqiang Meng","doi":"10.1016/j.asoc.2025.113087","DOIUrl":"10.1016/j.asoc.2025.113087","url":null,"abstract":"<div><div>Deep neural networks have achieved remarkable progress, but still face significant challenges such as domain shift and the scarcity of labeled data. Existing pseudo-labeling methods are affected by domain bias when dealing with distribution shifts across domains. To address these challenges, this paper proposes a novel self-adaptive framework for semi-supervised domain generalization. The proposed framework has three key components: (1) DGMatch, a pseudo-labeling method that encourages domain fairness and uncertainty weighting to generate more reliable pseudo-labels, thus overcoming the domain bias issues of existing pseudo-labeling approaches; (2) AutoMixLayer, a plug-and-play module that automatically adjusts the parameters for domain-style mixing, thus reducing the domain shift in the source data; and (3) DCT, a method that randomly transfers the target domain-style to a prototype of the source domain-style, effectively calibrating the domain-style during the test stage. These three modules allow the proposed framework to effectively alleviate the problem of performance degradation when faced with domain bias. Extensive experiments on image classification benchmarks and a medical image segmentation dataset demonstrate that the proposed self-adaptive framework outperforms state-of-the-art semi-supervised domain generalization methods. Further analysis reveals that the framework can effectively balance the quantity and quality of pseudo-labels, enabling robust model generalization.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113087"},"PeriodicalIF":7.2,"publicationDate":"2025-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143776298","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 multi-objective optimization based on knowledge prediction and density clustering strategy","authors":"Yong Wang,&nbsp;Shengao Wang,&nbsp;Kuichao Li,&nbsp;Gai-Ge Wang","doi":"10.1016/j.asoc.2025.113099","DOIUrl":"10.1016/j.asoc.2025.113099","url":null,"abstract":"<div><div>Dynamic multi-objective evolutionary algorithms (DMOEAs) that extract historical knowledge from the past environment to predict new solutions are known to be effective for solving dynamic multi-objective optimization problems (DMOPs). However, most of the existing methods simply reuse historical solutions without further extracting the knowledge between different historical environment solutions, which may make the algorithm ignore some important historical knowledge and limit its performance. In this paper, we propose a knowledge prediction strategy and a density clustering strategy for DMOEA, called KPDCS-DMOEA, which aim to extract historical knowledge from the past environment to build a more accurate prediction model. Firstly, the trend of change in the initial environment is obtained by predicting previous environmental changes through linear prediction methods based on knee point clusters. Secondly, a strategy was proposed to pair the solutions between adjacent environments and construct each dimensional motion vector as historical knowledge. The training set is constructed according to the motion step of the motion vector and the motion direction of each dimension, and the neural network is trained to predict the initial population in the new environment. Finally, a guided evolution strategy based on a density clustering algorithm is developed to speed up population convergence and ensure that the population is well distributed. KPDCS-DMOEA is compared with several state-of-the-art DMOEAs. Experimental results show that the performance of KPDCS-DMOEA is better than the selected comparison algorithms.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113099"},"PeriodicalIF":7.2,"publicationDate":"2025-03-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143776299","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":"Forecasting realized volatility using deep learning quantile function","authors":"Jungyoon Song ,&nbsp;Hyunju Lee ,&nbsp;Jongu Lee ,&nbsp;Woojin Chang","doi":"10.1016/j.asoc.2025.113016","DOIUrl":"10.1016/j.asoc.2025.113016","url":null,"abstract":"<div><div>The accurate prediction of realized volatility is an essential component of effective investment strategies. Existing studies have often focused on modeling selective features of intraday return series, overlooking the comprehensive information embedded within them due to challenges such as microstructure noise and the complexity of handling numerous data points. To address these limitations, this paper proposes a novel deep learning quantile function (DLQF) framework that directly leverages intraday return series to forecast realized volatility. The proposed model integrates a Bi-LSTM network to capture the long memory of realized volatility and a quantile function implemented as a deep neural network to extract rich information from intraday returns. A loss function based on <span><math><msub><mrow><mi>L</mi></mrow><mrow><mi>p</mi></mrow></msub></math></span> distance measures is defined to estimate the probabilistic distribution of intraday returns, enabling both intraday return prediction and realized volatility estimation. Empirical results demonstrate that DLQF outperforms traditional benchmarks across major ETFs, including SPY, DIA, and QQQ, which represent the S&amp;P 500, Dow Jones Industrial Average, and Nasdaq 100, respectively. This model offers significant potential for applications in portfolio optimization, option pricing, and risk management.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113016"},"PeriodicalIF":7.2,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143759721","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}