Applied Soft Computing最新文献

{"title":"Multi-view self-supervised learning on heterogeneous graphs for recommendation","authors":"Yunjia Zhang ,&nbsp;Yihao Zhang ,&nbsp;Weiwen Liao ,&nbsp;Xiaokang Li ,&nbsp;Xibin Wang","doi":"10.1016/j.asoc.2025.113056","DOIUrl":"10.1016/j.asoc.2025.113056","url":null,"abstract":"<div><div>Graph neural networks (GNNs) have significantly contributed to data mining but face challenges due to sparse graph data and lack of labels. Typically, GNNs rely on simple feature aggregation to leverage unlabeled information, neglecting the richness inherent in unlabeled data within graphs. Graph self-supervised learning methods effectively capitalize on unlabeled information. Nevertheless, most existing graph self-supervised learning methods focus on homogeneous graphs, ignoring the heterogeneity of graphs and mainly considering the graph structure from a single perspective. These methods cannot fully capture the complex semantics and correlations in heterogeneous graphs. It is challenging to design self-supervised learning tasks that can fully capture and represent complex relationships in heterogeneous graphs.</div><div>In order to address the above problems, we investigate the problem of self-supervised HGNN and propose a new self-supervised learning mechanism for HGNN called Multi-view Self-supervised Learning on Heterogeneous Graphs for Recommendation (MSRec). We introduce a maximum entropy path sampler to help sample meta-paths containing structural context. Encoding information from diverse views defined by various meta-paths, decoding it into a semantic space different from own and optimizing tasks in both local-view and global-view contrastive learning, which facilitates collaborative and mutually supervisory interactions between the two views, leveraging unlabeled information for node embedding learning effectively. According to experimental results, our method demonstrates an optimal performance improvement of approximately 7% in NDCG@10 and about 8% in Prec@10 compared to state-of-the-art models. The experimental results on three real-world datasets demonstrate the superior performance of MSRec compared to state-of-the-art recommendation methods.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 113056"},"PeriodicalIF":7.2,"publicationDate":"2025-03-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143724713","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":"Unraveling asset pricing with AI: A systematic literature review","authors":"Yan Chen ,&nbsp;Lin Zhang ,&nbsp;Zhilong Xie ,&nbsp;Wenjie Zhang ,&nbsp;Qing Li","doi":"10.1016/j.asoc.2025.112978","DOIUrl":"10.1016/j.asoc.2025.112978","url":null,"abstract":"<div><div>Asset pricing, long recognized as a cornerstone of financial studies with multiple Nobel Prizes in Economics, is experiencing a profound transformation through the integration of artificial intelligence (AI). This study highlights the convergence of finance and computer science in asset pricing, offering novel insights into AI-driven approaches through an in-depth analysis of hundreds of research papers. The study begins by examining the key factors influencing asset pricing, highlighting the significance of factor interactions in AI-driven asset pricing models. It then systematically reviews various econometric and machine learning models from both financial and computational perspectives, underscoring the importance of designing predictive asset pricing models based on financial assumptions and principles. This reflects the inevitable convergence of finance and computer science in the field of asset pricing. Finally, the study outlines three research directions, providing actionable guidance for future exploration: (1) the development of large-scale multimodal datasets to equip advanced models with the breadth of information required to enhance foresight, (2) the integration of fundamental economic theories into model design to enhance relevance and resilience, emulating the nuanced decision-making processes of experienced traders, and (3) improving the interpretability of deep learning models to bridge the gap between their outputs and actionable insights. In addition, this study introduces the <em>QuantPlus</em> project, an initiative designed to provide large-scale datasets that empower researchers to evaluate and advance innovative models.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 112978"},"PeriodicalIF":7.2,"publicationDate":"2025-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143768479","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":"Intra and inter-series pattern representations fusion network for multiple time series forecasting","authors":"Canghong Jin ,&nbsp;Tianyi Chen ,&nbsp;Hao Ni ,&nbsp;Qihao Shi","doi":"10.1016/j.asoc.2025.113024","DOIUrl":"10.1016/j.asoc.2025.113024","url":null,"abstract":"<div><div>Multiple time series (MTS) can comprise data collected from various wireless sensor networks in the actual application, and each source provides a distinctive pattern. Most existing neural network methods attempt to model the patterns of individual time series by training a global model using the entire dataset, suffering from insufficient ability to consider the differences among source patterns and lowering the predictability. To address this limitation, we propose the Multiple Time Series Pattern Representation Network(MTS-PRNet), a unified framework consisting of two modules to forecast multiple time series from diverse sources. The first is the intra-series correlation learning module, which explicitly learns the temporal dependencies of time series. The second is the inter-series discriminative representation learning module that learns shapelets as discriminative representations to capture shared features among series. By integrating the covariates map generated by the second module, both intra and inter-series characteristics are captured to provide transferable guidance for increasing predictability. Experiments conducted on 9 datasets verify that our model achieves state-of-the-art performance. In particular, we carry out an ablation study to validate the effectiveness of discriminative representations.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113024"},"PeriodicalIF":7.2,"publicationDate":"2025-03-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143724231","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 parameters genetic algorithm for collaborative strike task allocation of unmanned aerial vehicle clusters towards heterogeneous targets","authors":"Chao Zhang ,&nbsp;Jianlu Guo ,&nbsp;Fei Wang ,&nbsp;Boyuan Chen ,&nbsp;Chunshi Fan ,&nbsp;Linghui Yu ,&nbsp;Zhiwen Wang","doi":"10.1016/j.asoc.2025.113075","DOIUrl":"10.1016/j.asoc.2025.113075","url":null,"abstract":"<div><div>Collaborative strikes by unmanned aerial vehicle clusters (UAVCs) is becoming a key focus in the future air warfare, which can significantly enhance warfare effectiveness and reduce costs. To exactly describe the real battlefield scenarios, various heterogeneous strike-targets should be embedded. However, it will significantly increase the complexity of multi-constraint combinatorial optimization problem, thus the traditional genetic algorithm (GA) is difficult to solve efficiently due to its unchanged gene operator. In this paper, a dynamic parameters genetic algorithm has been proposed for UAVCs collaborative task allocation towards heterogeneous targets. Firstly, according to the differences of type, value, combat and defense, the heterogeneous strike-targets have been abstracted into strike target points and the UAVCs have been formulated into a set. Secondly, an innovative multiple unmanned aerial vehicles duplicate tasks orienteering problem (MUDTOP) model has been built to achieve multiple strikes on certain targets. Finally, the new triple-chromosome encoding and duplicate gene segments have been designed, and a novel genetic algorithm called DPGA-TEDG has been presented through dynamic gene operator. Experimental comparison results across various battlefield scales demonstrate that the outcomes of the proposed DPGA-TEDG algorithm not only meet practical requirements, but also outperform that of the other three algorithms in both optimality and robustness. Especially, in the battlefield scale environment of 180 km* 180 km, the average objective value of DPGA-TEDG is better than that of traditional genetic algorithm (GA-TEDG), simulated annealing algorithm (SA) and particle swarm optimization algorithm (PSO) about 2.71 %, 6.58 % and 20.49 %, respectively.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113075"},"PeriodicalIF":7.2,"publicationDate":"2025-03-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143724090","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":"Railway prioritized food logistics in developing countries using fuzzy decision making under interval-valued pythagorean fuzzy environment","authors":"Ali Atilla Arisoy ,&nbsp;S. Jeevaraj ,&nbsp;Ilgin Gokasar ,&nbsp;Muhammet Deveci ,&nbsp;Seifedine Kadry ,&nbsp;Zhe Liu","doi":"10.1016/j.asoc.2025.113066","DOIUrl":"10.1016/j.asoc.2025.113066","url":null,"abstract":"<div><div>The current state of agricultural logistics is vulnerable to global crises and oil price fluctuations, especially in developing countries that depend heavily on highway transportation. Experts are seeking efficient and eco-friendly solutions, exploring options such as railroad transport and innovative concepts such as synchromodality for improvement. In this study, a decision-making approach for policymakers and logistics experts to improve the efficiency and resilience of agricultural logistics by using more sustainable transport modes and synchromodality is proposed. The approach is based on a new total ordering principle on the class of Interval-Valued Pythagorean Fuzzy Numbers (IVPFNs), which is compared with existing ranking methods. In this paper, we have used IVPFNs for modelling our problem. The idea of IVPFNs (generalising interval-valued intuitionistic fuzzy numbers) introduced by Yager in 2013. However, the total ordering of the class of IVPFNs has not been studied so far. The main Mathematical contribution of this work lies in defining the total order relation on the set of IVPFNs for the first time in the literature. To do this, firstly, the Four new score functions on the set of IVPFNs are introduced and various mathematical properties of them are studied. Secondly, a new total ordering principle is introduced by combining all these score functions, and their mathematical proofs are given. Thirdly, a new group decision-making algorithm based on interval-valued Pythagorean fuzzy extent analysis (IVPFEA) is proposed and applied to a real-life case study problem. Finally, the sensitivity analysis has been done properly to show the robustness of the proposed algorithm and the results. The case study involves seven experts role-playing as advisors for the Republic of Türkiye, which is a developing country, on choosing the best agricultural logistics system alternative among four alternatives. Twelve criteria, under four aspects, are presented for participants to consider. Based on the responses of the experts, the railway-prioritized food logistics system is the primary alternative. Overall, the results of this study provide a mathematical and data-driven approach to deciding on a new logistics system that policymakers and sector experts can utilize.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113066"},"PeriodicalIF":7.2,"publicationDate":"2025-03-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143738287","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":"FT-GPNN: A finite-time convergence solution for multi-set constrained optimization","authors":"Huiting He ,&nbsp;Chengze Jiang ,&nbsp;Zhiyuan Song ,&nbsp;Xiuchun Xiao ,&nbsp;Neal Xiong","doi":"10.1016/j.asoc.2025.113030","DOIUrl":"10.1016/j.asoc.2025.113030","url":null,"abstract":"<div><div>Gradient Neural Networks (GNNs) have demonstrated remarkable progress in handling optimization problems. However, applying GNNs to multi-constrained optimization problems, particularly those with those involving multi-set constraints, poses several challenges. These challenges arise from the complexity of the derivations and the increasing number of constraints. As the number of constraints increases, the optimization problem becomes more complex, making it more challenging for GNN-based methods to effectively identify the optimal solution. Motivated by these challenges, the Finite-Time Gradient Projection Neural Network (FT-GPNN) is introduced for tackling Multi-set Constrained Optimization (MCO). This innovative solution incorporates an Enhanced Sign-Bi-Power (ESBP) activation function and simplifies the design tailored explicitly for MCO. Furthermore, within the Lyapunov stability framework, the theoretical foundation of this model is strengthened by rigorous proof of local convergence. Building upon this foundation, we further establish that our model can achieve convergence within a finite time. To validate the effectiveness of our approach, we present empirical results from numerical experiments conducted under consistent conditions. Notably, our experiments demonstrate that the model using the ESBP activation function outperforms others in terms of finite-time convergence.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113030"},"PeriodicalIF":7.2,"publicationDate":"2025-03-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143724071","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":"Thinking Innovation Strategy (TIS): A novel mechanism for metaheuristic algorithm design and evolutionary update","authors":"Heming Jia,&nbsp;Xuelian Zhou,&nbsp;Jinrui Zhang","doi":"10.1016/j.asoc.2025.113071","DOIUrl":"10.1016/j.asoc.2025.113071","url":null,"abstract":"<div><div>The metaheuristic optimization algorithm(MHS) is a global optimization method inspired by natural phenomena, demonstrating superior performance in specific application scenarios. Traditional optimization algorithms utilize two main concepts: exploration, to expand the search range, and exploitation, to enhance solution accuracy. However, as problem complexity and application scenarios increase, MHS struggles to balance exploration and exploitation to find the optimal solution. Therefore, this paper introduces innovative characteristics of individual thinking and proposes a new Thinking Innovation Strategy (TIS). TIS does not aim for an optimal solution but seeks global optimization based on successful individuals, enhancing algorithm performance through survival of the fittest. This paper applies TIS strategies to improve various MHS algorithms and evaluates their performance on 57 engineering problems and the IEEE CEC2020 benchmarks. Experimental results indicate that the TIS-enhanced algorithms outperform the original versions across 57 engineering problems, according to Friedman ranking and Wilcoxon rank-sum test results. Some algorithms show significant improvement, demonstrating the feasibility and practicality of TIS for optimization problems. The TIS (LSHADE_SPACMA) of the source code can be accessed through the following ways: https://github.com/LIANLIAN-Serendipity/TIS-</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113071"},"PeriodicalIF":7.2,"publicationDate":"2025-03-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143724233","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":"Nested deep learning with learned network embeddings for software defect prediction","authors":"Sweta Mehta ,&nbsp;Lov Kumar ,&nbsp;Sanjay Misra ,&nbsp;K.Sridhar Patnaik ,&nbsp;Vikram Singh","doi":"10.1016/j.asoc.2025.113057","DOIUrl":"10.1016/j.asoc.2025.113057","url":null,"abstract":"<div><div>Existing software (SW) defect prediction approaches and the models are majorly based on features extracted from the code of the software to build defect datasets for predictive modeling. However, these models fail to sufficiently capture the complex, latent dependencies within the software components, which acts as a hindrance in achieving higher predictive accuracy. This study introduces an improved defect prediction model, the Nested Deep Learning (NDL) model, that leverages network embeddings from call graphs for enhanced representation of intricate hierarchical class dependencies and interactions. This work evaluates six network-embedding algorithms by applying them to call graphs of 10 real software projects, generating embeddings of dimensions 32 and 128. A total of 50 NDL models—with and without dropout layers—are developed, and a comparative evaluation of these models is conducted against traditional classifier-based models. This evaluation demonstrated the superiority of the NDL model with dropout, achieving a mean AUC of 0.87, an 8.98 % improvement over the traditional classifier-based models. Among the evaluated embedding methods, LINE embeddings outperformed others, and integrating network embeddings with software metrics led to a 15.85 % AUC improvement over using software metrics alone. The optimal configuration—combining software metrics with LINE embeddings (dimension 128) in an NDL model with three deep learning layers and dropout—achieved a mean AUC of 0.93, surpassing all other configurations by 3.33–14.81 %<strong>.</strong> This study is the first to validate the effectiveness of a nested deep learning framework for modeling call graph dependencies through network embeddings, providing a scalable and robust approach for improving software defect prediction.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 113057"},"PeriodicalIF":7.2,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143724712","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":"Cancelable binary biometric template generation scheme based on partial walsh transformation and MinHash algorithm","authors":"Shuaichao Song,&nbsp;Yeming Yang,&nbsp;Miao Yu,&nbsp;Yuming Liao,&nbsp;Weilai Guo,&nbsp;Jiyuan Li,&nbsp;Songhui Guo","doi":"10.1016/j.asoc.2025.113049","DOIUrl":"10.1016/j.asoc.2025.113049","url":null,"abstract":"<div><div>With the widespread use of biometrics, biometric templates stored in biometric systems are at serious risk of security and privacy breaches. Cancelable biometric scheme is an effective remedy when many unprotected biometric templates are compromised. We propose a cancelable binary biometric template generation scheme based on the partial Walsh transformation and the MinHash algorithm to improve recognition accuracy and generation efficiency. Firstly, the partial Walsh matrix transforms the high-dimensional original biometric feature into a low-dimensional space. Then, protected cancelable binary biometric templates are generated based on the proposed sliding window grouping minimum hash algorithm SWG-MinHash. Our scheme demonstrates superior recognition accuracy and generation efficiency on fingerprint and face databases compared to existing schemes. Meanwhile, our scheme satisfies the properties of non-invertibility, revocability, and unlinkability, and is resistant to common security and privacy attacks. Therefore, our scheme effectively mitigates the problem of balancing recognition accuracy, security, and generation efficiency of cancelable biometric schemes and is more practical for biometric systems. The source code of our scheme is available at <span><span>https://github.com/sscwrx/cbef</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113049"},"PeriodicalIF":7.2,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143724229","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 performance-driven multi-stage KNN approach for local adaptive classification","authors":"Che Xu ,&nbsp;Zhenhua Fan","doi":"10.1016/j.asoc.2025.113070","DOIUrl":"10.1016/j.asoc.2025.113070","url":null,"abstract":"<div><div>A key issue of the K-Nearest Neighbors (KNN) algorithm is determining the optimal neighborhood size <em>K</em>, which limits the widespread applicability of KNN. To address this, a performance-driven multi-stage KNN (PMKNN) approach is proposed in this paper. Given a set of alternative <em>K</em> values, the traditional KNN algorithm is initially employed in the PMKNN approach to identify the optimal <em>K</em> values for all known samples. A convex optimization model is then constructed based on the least squares loss function to learn the correlation between known samples and query samples. After the learned correlation is used to evaluate the performances of all candidate <em>K</em> values in classifying query samples, a weighted majority voting process is designed to generate the final classification results. Unlike existing KNN approaches, the proposed PMKNN approach considers multiple optimal <em>K</em> values for each query sample, enhancing classification stability and reliability. The proposed approach also reduces the negative impact of inappropriate <em>K</em> values on classification performance. An experimental study is conducted using twenty real-world classification datasets collected from two public data repositories to assess the effectiveness of the proposed PMKNN approach. The relevant results highlight the high classification performance of the proposed PMKNN approach compared to seven state-of-the-art KNN methods and underscore its predictive stability compared to the traditional KNN algorithm using all possible <em>K</em> values.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113070"},"PeriodicalIF":7.2,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143724091","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}