{"title":"Developing a forecasting model for time series based on clustering and deep learning algorithms","authors":"Luan Nguyen-Huynh , Tai Vo-Van","doi":"10.1016/j.asoc.2025.112977","DOIUrl":"10.1016/j.asoc.2025.112977","url":null,"abstract":"<div><div>This study proposes a new forecasting model for time series based on the improvement and combination of the cluster analysis (CA) algorithm and deep learning with Convolutional Neural Network (CNN) and Bi-Long Short Term Memory (BiLSTM) model. The proposed model is considered pioneering in this research direction with significant contributions to three main phases. For the first phase, the original series is converted into the percentage change series and is divided into clusters of an appropriate number using the CA algorithm. The next phase involves extracting the features of the new series based on the CNN with suitable parameters and input data enhancement from the results of the first phase. In the final phase, the BiLSTM model is applied to the series established from the second phase, and the forecasting principle for the future is established. The proposed model is detailed in the implementation steps, proving convergence, illustrated by numerical examples, and can be applied to real series using a Matlab procedure. The effectiveness of the proposed model is quite impressive as it surpasses many strong forecasting models on reputable benchmark datasets , including the M3-Competition dataset with 3,003 series, and M4-Competition dataset with 100,000 series.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 112977"},"PeriodicalIF":7.2,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143643859","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":"Optimized echo state network for error compensation based on transfer learning","authors":"Yingqin Zhu , Yue Liu , Zhaozhao Zhang , Wen Yu","doi":"10.1016/j.asoc.2025.112935","DOIUrl":"10.1016/j.asoc.2025.112935","url":null,"abstract":"<div><div>Echo State Network (ESN) is widely applied in nonlinear system modeling, but its performance is often limited by a lack of error autocorrelation analysis, leading to reduced modeling accuracy. Existing extensions, such as SR-ESN and ERBM, primarily focus on structural optimization or feature representation but fail to effectively address autocorrelation errors. To overcome these limitations, we propose a Transfer Learning-based Echo State Network (TLESN) that compensates for errors in realtime to enhance prediction accuracy. The TLESN integrates a computing layer based on ESN and a compensation layer employing transfer learning, which dynamically adjusts output weights. To validate the proposed model, experiments are conducted on the Mackey-Glass time series, a practical Sunspot dataset, and a real-world industrial dataset. Results demonstrate that TLESN effectively mitigates autocorrelation errors, achieving at least a 17% improvement in prediction accuracy compared to existing ESN extensions.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 112935"},"PeriodicalIF":7.2,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143687353","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}
Xiaoxu Guo , Siyan Liang , Yachao Cui , Juxiang Zhou , Lei Wang , Han Cao
{"title":"Multimodal fine-grained reasoning for post quality evaluation","authors":"Xiaoxu Guo , Siyan Liang , Yachao Cui , Juxiang Zhou , Lei Wang , Han Cao","doi":"10.1016/j.asoc.2025.112955","DOIUrl":"10.1016/j.asoc.2025.112955","url":null,"abstract":"<div><div>Accurate assessment of post quality frequently necessitates complex relational reasoning skills that emulate human cognitive processes, thereby requiring the modeling of nuanced relationships. However, existing research on post-quality assessment suffers from the following problems: (1) They are often categorization tasks that rely solely on unimodal data, which inadequately captures information in multimodal contexts and fails to differentiate the quality of students’ posts finely. (2) They ignore the noise in the multimodal deep fusion between posts and topics, which may produce misleading information for the model. (3) They do not adequately capture the complex and fine-grained relationships between post and topic, resulting in an inaccurate evaluation, such as relevance and comprehensiveness. Based on the above challenges, the Multimodal Fine-grained Topic-post Relational Reasoning(MFTRR) framework is proposed for modeling fine-grained cues by simulating the human thinking process. It consists of the local–global semantic correlation reasoning module and the multi-level evidential relational reasoning module. Specifically, MFTRR addresses the challenge of unimodal and categorization task limitations by framing post-quality assessment as a ranking task and integrating multimodal data to more effectively distinguish quality differences. To capture the most relevant semantic relationships, the Local–Global Semantic Correlation Reasoning Module enables deep interactions between posts and topics at both local and global scales. It is complemented by a topic-based maximum information fusion mechanism to filter out noise. Furthermore, to model complex and subtle relational reasoning, the Multi-Level Evidential Relational Reasoning Module analyzes topic-post relationships at both macro and micro levels by identifying critical cues and delving into granular relational cues. MFTRR is evaluated using three newly curated multimodal topic-post datasets, in addition to the publicly available Lazada-Home dataset. Experimental results indicate that MFTRR outperforms state-of-the-art baselines, achieving a 9.52% improvement in the NDCG@3 metric compared to the best text-only method on the Art History course dataset.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 112955"},"PeriodicalIF":7.2,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143643856","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 crude oil price forecasting framework based on Constraint Guarantee and Pareto Fronts Shrinking Strategy","authors":"Yujie Chen , Zhirui Tian","doi":"10.1016/j.asoc.2025.112996","DOIUrl":"10.1016/j.asoc.2025.112996","url":null,"abstract":"<div><div>Accurate forecasting of crude oil prices is essential for making informed energy policy decisions and ensuring energy security. However, crude oil price forecasting is inherently challenging due to the volatile, nonlinear, and complex nature of the market. While ensemble learning approaches have shown promise in enhancing forecasting accuracy, many existing models rely on multi-objective optimization techniques that generate a Pareto frontier of optimal solutions, often making it difficult to select the best solution for practical application. This issue is exacerbated by the fact that some Pareto-optimal solutions are not suitable for real-world decision-making, leading to inefficiencies in model performance. To address these limitations, this research proposes a novel ensemble learning framework that incorporates a Constraint Guarantee Strategy (CGS) and a Pareto Front Shrinking Strategy (PFSS) to enhance both the accuracy and stability of crude oil price forecasting models. The CGS filters out inferior solutions during the optimization process, ensuring that the ensemble model outperforms individual models in terms of forecasting accuracy. The PFSS helps decision-makers select the most relevant solutions from the Pareto frontier by balancing trade-offs between objectives and narrowing down the set of solutions. Our framework is evaluated on three widely used datasets: Brent, WTI, and Dubai crude oil prices, and compared with state-of-the-art models from both the general time-series forecasting domain and crude oil price forecasting. It improves prediction accuracy by approximately 23.2% on the Brent dataset, 4.0% on the WTI dataset, and 21.7% on the Dubai dataset, based on improvements in MAPE. Ablation studies confirm the effectiveness of each component. The discussion further emphasizes the practical applicability and robustness of the framework, confirming its potential for real-world crude oil price forecasting.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 112996"},"PeriodicalIF":7.2,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143686927","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}
Jing Chen , Haocheng Ye , Zhian Ying , Yuntao Sun , Wenqiang Xu
{"title":"Dynamic trend fusion module for traffic flow prediction","authors":"Jing Chen , Haocheng Ye , Zhian Ying , Yuntao Sun , Wenqiang Xu","doi":"10.1016/j.asoc.2025.112979","DOIUrl":"10.1016/j.asoc.2025.112979","url":null,"abstract":"<div><div>Accurate traffic flow prediction is essential for applications like transport logistics but remains challenging due to complex spatio-temporal correlations and non-linear traffic patterns. Existing methods often model spatial and temporal dependencies separately, failing to effectively fuse them. To overcome this limitation, the <strong>D</strong>ynamic <strong>S</strong>patial-<strong>T</strong>emporal <strong>T</strong>rend Trans<strong>former</strong> (<strong>DST<sup>2</sup>former</strong>) is proposed to capture spatio-temporal correlations through adaptive embedding and to fuse dynamic and static information for learning multi-view dynamic features of traffic networks. The approach employs the <strong>D</strong>ynamic <strong>T</strong>rend <strong>R</strong>epresentation Trans<strong>former</strong> (<strong>DTRformer</strong>) to generate dynamic trends using encoders for both temporal and spatial dimensions, fused via Cross Spatial-Temporal Attention. Predefined graphs are compressed into a representation graph to extract static attributes and reduce redundancy. Experiments on four real-world traffic datasets demonstrate that our framework achieves state-of-the-art performance.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 112979"},"PeriodicalIF":7.2,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143642507","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":"Attention-aware graph contrastive learning with topological relationship for recommendation","authors":"Xian Mo , Jun Pang , Zihang Zhao","doi":"10.1016/j.asoc.2025.113008","DOIUrl":"10.1016/j.asoc.2025.113008","url":null,"abstract":"<div><div>Recommender systems are a vital tool to guide the overwhelming amount of online information for users, which has been successfully applied to online retail platforms, social networks, etc. Recently, contrastive learning has revealed outstanding performance in recommendation by data augmentation strategies to handle highly sparse data. Most existing work fails to leverage the original network’s topology to construct attention-aware modules that identify user–item interaction importance for guiding node aggregation while preserving key semantics and reducing noise in the reconstructed graph during data augmentation. In this paper, our work proposes an <u>At</u>t<u>e</u>ntion-aware <u>G</u>raph <u>C</u>ontrastive <u>L</u>earning architecture with Topological Relationship (AteGCL) for recommendation. In particular, our AteGCL proposes an attention-aware mechanism with topological relationships to learn the importance between users and items for extracting the local graph dependency, which identifies the importance between nodes by constructing an attention-aware matrix into graph convolutional networks using a random walk with a restart strategy for generating node feature aggregation. We then employ principal component analysis (PCA) for contrastive augmentation and utilize the attention-aware matrix to ease noise from the reconstructed graph generated by PCA and to generate a new view with global collaborative relationships and less noise. Comprehensive experiments on three real-world user–item networks reveal the superiority of our AteGCL over diverse state-of-the-art recommendation approaches. Our code is available at <span><span>https://github.com/ZZHCodeZera/AteGCL</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 113008"},"PeriodicalIF":7.2,"publicationDate":"2025-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143642574","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 Fuzzy AHP-based trust management mechanism for self-sovereign identity in the metaverse","authors":"Xiaoling Song , Guangxia Xu , Yongfei Huang","doi":"10.1016/j.asoc.2025.112994","DOIUrl":"10.1016/j.asoc.2025.112994","url":null,"abstract":"<div><div>Self-sovereign identity (SSI) technology has advantages and potential for application in the metaverse. However, the decentralization and anonymous interaction of SSI create convenience for malicious attacks, frauds, and conspiracies in the metaverse. It leads to various trust risks and threats to the meta-universe system. To address these challenges, we analyze the risks of SSI systems and constructed a reputation index system. Moreover, we propose a blockchain-based reputation management framework (BBRMF), which can constrain users from engaging in illegal activities such as forgery, fraud, and conspiracy, thereby guaranteeing the security and trustworthiness of the entities involved in the metaverse. In BBRMF, we constructed a reputation evaluation model based on fuzzy analytical hierarchy process (FAHP) to assess the user’s reputation in three dimensions: reliability, trustworthiness and security. To motivate users to accumulate more positive reputation, we set the user’s reputation score into a reputation credential in the form of non-fungible token (NFT), through which users can obtain more benefits and opportunities. Finally, we calculated the reputation value of SSI related entities from multiple perspectives through simulation experiments and comparative analysis. The feasibility of the proposed method is verified, and it is proved that it can effectively resist the interference and attack of malicious scoring nodes. Moreover, the scheme adopts multi-dimensional evaluation indexes and behavioral feature values, which significantly improves the comprehensiveness and accuracy of the reputation assessment. Meanwhile, the weights of the evaluation indexes are derived through objective calculation, ensuring the fairness of the evaluation results, and improving the credibility and repeatability of the reputation assessment.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 112994"},"PeriodicalIF":7.2,"publicationDate":"2025-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143643854","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":"Short-term forecasting of electricity price using ensemble deep kernel based random vector functional link network","authors":"Someswari Perla , Ranjeeta Bisoi , P.K. Dash , A.K. Rout","doi":"10.1016/j.asoc.2025.113012","DOIUrl":"10.1016/j.asoc.2025.113012","url":null,"abstract":"<div><div>Accurate short-term electricity price forecasting in a deregulated electrical market is a difficult task as the electricity price exhibits high nonlinearity, sharp price spikes, and seasonality in different frequencies, etc. Thus, this study presents a new approach using an Ensemble Deep Kernel Random Vector Functional Link Network (EDKRVFLN) model hybridized with a Chaotic Sine Cosine Improved Firefly Algorithm (CSCIFA) for short-term electricity price forecasting with better generalization capacity, simple structure, and significant accuracy. Unlike the Ensemble Deep Random Vector Functional Link Network (EDRVFLN) where each stacked layer requires proper choice of the number of hidden nodes and manual tuning of random weights and biases along with the pseudoinverse solution of the output weights in each layer leading to suboptimal model generalization. However, the choice of random weights and biases along with the number of hidden neurons in the proposed EDKRVFLN model can be dispensed by using kernel-based transformation and representation learning. Further each stacked layer of the proposed model utilizes kernel based linear features from the direct links and nonlinearly transformed features from the enhancement nodes from the preceding layers of the prediction model. Also, each layer produces an output by simple invertible kernel matrix inversion based on generalized least squares, and the final output is the ensemble of the outputs from each layer, thus simultaneously producing an ensemble and deep learning framework. Seven electricity price datasets are examined to confirm the supremacy of the proposed model in comparison to several benchmark models.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 113012"},"PeriodicalIF":7.2,"publicationDate":"2025-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143686926","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 , Xiao Wu , Jiaying Zhang , Weiping Wang , Linjiang Zheng , Jiaxing Shang
{"title":"Series clustering and dynamic periodic patching-based transformer for multivariate time series forecasting","authors":"Yijie Wang , Xiao Wu , Jiaying Zhang , Weiping Wang , Linjiang Zheng , Jiaxing Shang","doi":"10.1016/j.asoc.2025.112980","DOIUrl":"10.1016/j.asoc.2025.112980","url":null,"abstract":"<div><div>Multivariate time series forecasting (MTSF) is widely employed in research-intensive domains, such as weather forecasting. Recently, Transformer-based models have outstanding ability to achieve SOTA performance, benefiting from its self-attention mechanism. However, existing models fall short in capturing multivariate inter-dependencies and local semantic representations. To tackle the above limitations, we propose a series clustering and dynamic periodic patching-based Transformer model named CMDPPformer, with two distinctive characteristics: (1) A channel-mixing module based on series clustering is proposed which can strengthen the association between variables with high sequence similarity, and weaken the effect of uncorrelated variables. Concretely, we use whole-time series clustering to group multivariate time series into clusters. After that, variables in the same cluster share the same Transformer backbone while variables in different clusters do not affect each other. (2) A dynamic periodic patching module is introduced which can better capture semantic information and improve Transformer’s local semantic representation. Concretely, multivariate time series after clustering are dynamically segmented into periodic patches as Transformer’s input token. Experimental results show that CMDPPformer can achieve an overall 13.76% and 10.16% relative improvements than SOTA Transformer-based models on seven benchmarks, covering four real-world applications: energy, weather, illness and economic.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 112980"},"PeriodicalIF":7.2,"publicationDate":"2025-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143629469","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}
Huanyi Ye , Jiale Guo , Ziyao Liu , Yu Jiang , Kwok-Yan Lam
{"title":"Enhancing AI safety of machine unlearning for ensembled models","authors":"Huanyi Ye , Jiale Guo , Ziyao Liu , Yu Jiang , Kwok-Yan Lam","doi":"10.1016/j.asoc.2025.113011","DOIUrl":"10.1016/j.asoc.2025.113011","url":null,"abstract":"<div><div>Recently, machine unlearning (MU) has received significant attention for its ability to remove specific undesired knowledge from a trained model, thereby ensuring AI safety. Furthermore, efforts have been made to integrate MU into existing Machine Learning as a Service (MLaaS), allowing users to raise requests to remove the influence of their data used in the training phase, after which the server conducts MU to remove its influence based on the unlearning requests. However, previous research reveals that malicious users may manipulate the requests so that the model utility may be significantly compromised after unlearning, which is known as malicious unlearning. In addition, privacy leakage may be exploited by malicious users by analyzing inference results obtained from the original model and the unlearned model. In this connection, we investigate these potential risks, specifically in ensemble models, which are widely adopted in MU because of their efficiency in unlearning and robustness in learning. However, despite these advantages, their vulnerabilities to malicious unlearning and privacy leakage remain largely unexplored. Our work explores malicious unlearning and malicious inference in ensemble settings. We propose a method in which malicious unlearning requests can trigger hidden poisons in ensembles, causing target images to be misclassified as intended by adversaries. Additionally, we introduce a privacy leakage attack where adversaries with black-box access to voting outputs can infer the unlearned label by analyzing the differences between the original and unlearned ensemble outputs. Experimental results demonstrate that these attacks can be highly stealthy and achieve a high success rate. Furthermore, comparative experiments reveal that these attacks present slightly lower stealthiness in ensemble settings compared to single-model scenarios, suggesting that ensemble models have advantages in detecting such malicious activities. These findings reveal that ensemble models are vulnerable to malicious unlearning and privacy leakage and highlight the urgent need for more robust MU designs to ensure AI safety.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"174 ","pages":"Article 113011"},"PeriodicalIF":7.2,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143687354","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}