{"title":"Geological and engineering insights from transfer learning with fourier neural operators: A case study of CO2 storage forecasting in disparate saline aquifers","authors":"Yusuf Falola , Siddharth Misra , Andres Nunez","doi":"10.1016/j.asoc.2025.113272","DOIUrl":"10.1016/j.asoc.2025.113272","url":null,"abstract":"<div><div>The rapid advancement of machine learning techniques, particularly Fourier Neural Operators (FNO), offers a promising approach to predicting CO<sub>2</sub> saturation and pressure distributions in geological carbon storage. This study explores the application of FNO combined with transfer learning (FNO+TL) to enhance computational efficiency and accuracy in forecasting CO<sub>2</sub> storage under diverse geological and operational conditions. We trained FNO models on datasets from the SACROC (153 samples) geological model and applied TL to predict outcomes for the Illinois Basin - Decatur Project (IBDP). Our findings highlight the substantial computational savings without significant compromise in performance of the FNO+TL models compared to FNO, using 10 and 20 samples for pressure and saturation predictions respectively. The FNO+TL model achieved an average Mean Absolute Error (MAE) of 0.11 for CO2 saturation and 8.7 psia for pressure predictions, compared to 0.79 and 2.4 psia respectively for FNO. While saturation predictions were less precise, the model effectively captured the overall CO<sub>2</sub> migration trends. Notably, transfer learning significantly reduced computational costs, decreasing training time by 62.5 % and storage, RAM requirements by 90 % and 68 %, respectively. Despite some limitations in saturation prediction accuracy, the FNO+TL approach demonstrates potential for efficient and reliable CO<sub>2</sub> storage forecasting. This study highlights the potential of FNOs and transfer learning for efficient and accurate forecasting of CO<sub>2</sub> storage behavior and management of carbon sequestration projects.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"178 ","pages":"Article 113272"},"PeriodicalIF":7.2,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144134083","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}
İlyas Karasu , Beyza Görkemli̇ Bayram , Mustafa Serdar Genç
{"title":"Predicting airfoil separation bubble locations using ABCP algorithms","authors":"İlyas Karasu , Beyza Görkemli̇ Bayram , Mustafa Serdar Genç","doi":"10.1016/j.asoc.2025.113309","DOIUrl":"10.1016/j.asoc.2025.113309","url":null,"abstract":"<div><div>In this study, new equations to predict the location parameters of the laminar separation bubble that occur in the flow over the blade/wing and negatively affect the blade/wing aerodynamic performance in unmanned aerial vehicles and wind turbines were developed first in the literature by Artificial Bee Colony Programming (ABCP) and quick ABCP (qABCP) algorithms. Data from the experimental study for NACA2415 were processed using ABCP and qABCP methods. The results of the models were also compared with the results of the XFOIL code, a numerical analysis in the literature, and an Artificial Neural Network (ANN). Even though low Reynolds numbers with more viscous effects were not given in the training data, both ABCP and qABCP algorithms successfully estimated the separation (Xs) and the reattachment points (Xr). Considering the error analysis and correlation coefficient values, it was seen that both algorithms can be used for both Xs and Xr predictions. Users/designers of the aerospace and energy industry can use to estimate Xr and Xs points for the NACA 2415 airfoil using the new equations proposed in this study at Re numbers ranging from 50,000 to 300,000, without the need for expensive and time-consuming experiments or Computational Fluid Dynamics (CFD) analysis. Furthermore, it was concluded that ABCP methods not only have the advantage of flexibly building models but are also highly competitive with other machine learning methods used in the literature for prediction, such as ANN.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"178 ","pages":"Article 113309"},"PeriodicalIF":7.2,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144154938","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":"Integrating multi-dimensional graph attention networks and transformer architecture for predicting air pollution in subway stations","authors":"Dingya Chen, Hui Liu","doi":"10.1016/j.asoc.2025.113033","DOIUrl":"10.1016/j.asoc.2025.113033","url":null,"abstract":"<div><div>Accurate prediction of PM<sub>2.5</sub> concentrations in subway stations is crucial for developing effective air pollution control strategies. However, existing methods struggle to accurately predict PM<sub>2.5</sub> concentrations due to the challenges of multi-step ahead forecasting, modeling long time series, capturing complex spatiotemporal correlations, and handling data quality issues such as missing values. This study proposes MAGICFormer, a novel hybrid end-to-end model for predicting PM<sub>2.5</sub> concentrations. The model comprises key components such as data preprocessing, a multi-dimensional graph attention network (md-GAT) module, as well as an Informer encoder and a Cross Decoder based on the Transformer architecture. The data preprocessing method improves data quality by addressing missing values and correcting anomalies. MAGICFormer integrates spatiotemporal correlations to predict PM<sub>2.5</sub> concentrations. The md-GAT module adaptively captures complex spatial relationships among subway stations across different dimensions, with its output serving as input to the Spatio Decoder. The Informer encoder processes long sequences and extracts temporal features, which are then passed to the Spatio Decoder and Temporal Decoder within the Cross Decoder for information fusion. The Cross Decoder aggregates the outputs of the Spatio and Temporal Decoders using a cross-attention mechanism, leveraging the interdependencies between graph-structured and time-series data to enhance prediction accuracy and improve model performance by effectively fusing spatial and temporal information. Experiments on Seoul subway stations show that MAGICFormer improves prediction accuracy by over 20 % compared to existing methods, demonstrating its effectiveness in long-term PM<sub>2.5</sub> forecasting. The proposed model offers a practical decision support tool for enhancing air quality management strategies in subway systems, particularly for long-term monitoring and control.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"178 ","pages":"Article 113033"},"PeriodicalIF":7.2,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144134080","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":"Ensemble self-organizing recursive neural network for modeling furnace temperature in municipal solid waste incineration","authors":"Tao Yu , Haixu Ding , Junfei Qiao","doi":"10.1016/j.asoc.2025.113170","DOIUrl":"10.1016/j.asoc.2025.113170","url":null,"abstract":"<div><div>The modeling of furnace temperature (FT) is the foundation of optimizing and controlling in municipal solid waste incineration (MSWI) process. However, owing to the high nonlinearity and dynamicity, complex reaction mechanisms, and strong coupling phenomena of MSWI process, accurately modeling the FT remains a significant challenge. In this paper, an ensemble self-organizing recursive neural network with information fusion gain algorithm (ESORNN-IFG) is proposed for FT modeling in MSWI process. First, the AdaBoost algorithm is introduced to combine multiple base learners by weighting them to enhance the accuracy and robustness of the strong learner. Second, an IFG index is introduced to appraise the contribution of hidden neurons and their interrelationships. Third, a self-organizing strategy combined with the IFG index is designed for adjusting the structure of the base learners during model training. Finally, the merits and effectiveness of the proposed ESORNN-IFG is confirmed by comparison with other existing approaches after testing the experimental results on several benchmark problems and the practical application of FT modeling in MSWI process.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"178 ","pages":"Article 113170"},"PeriodicalIF":7.2,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144124060","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":"Enhancing financial time series forecasting with hybrid Deep Learning: CEEMDAN-Informer-LSTM model","authors":"Jiang-Cheng Li, Li-Ping Sun, Xiao Wu, Chen Tao","doi":"10.1016/j.asoc.2025.113241","DOIUrl":"10.1016/j.asoc.2025.113241","url":null,"abstract":"<div><div>Financial time series forecasting is fraught with challenges due to the significant noise and uncertainty in the financial market that can bias model prediction outcomes. However, deep learning, as an important branch of artificial intelligence, has demonstrated a strong ability in dealing with large-scale nonlinear data. Therefore, in order to solve this problem, this paper mainly focuses on the closing price of the CSI 300 index as the research object and proposes a new deep learning hybrid prediction model, CEEMDAN-Informer-LSTM. The model decomposes the signals using complete ensemble empirical mode decomposition of adaptive noise (CEEMDAN) and classifies the decomposed signals by the zero-mean T-hypothesis testing method into high-frequency (H-IMF) and low-frequency components (L-IMF). In order to utilize the prediction advantages of different models in different frequency ranges, as well as to more accurately capture the intrinsic patterns and features in the signals, the combination of Informer prediction of H-IMF and Long Short Memory Model (LSTM) prediction of L-IMF is used to form a hybrid model. In the empirical study, this paper is presented by comparing with BP, RNN, LSTM, Informer, Transformer, iTransformer, CEEMDAN-BP, CEEMDAN-RNN, CEEMDAN-LSTM, CEEMDAN-Informer, CEEMDAN-Transformer, CEEMDAN-iTransformer models are compared and four loss functions, MAE, RMSE, MAPE, and <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span>, are selected to evaluate the model performance. The experimental results show that Informer has the best results in individual model prediction accuracy, followed by iTransformer, Transformer, LSTM, BP, and RNN. The prediction performance of the proposed CEEMDAN-Informer-LSTM hybrid prediction model is higher than all other models. In order to verify the robustness of the model, this paper has found out that the proposed hybrid model has good prediction accuracy and also exploited the application of the model in the stock market through multi-step prediction, MCS confidence test, dataset discussion, data leakage processing, replacing experimental data, and constructing a simple quantitative trading investment strategy from which the proposed hybrid model has good prediction accuracy.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"177 ","pages":"Article 113241"},"PeriodicalIF":7.2,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144084719","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}
Darian H. Grass-Boada, Leandro González-Montesino, Rubén Armañanzas
{"title":"Community detection in networks: A rough sets and consensus clustering approach","authors":"Darian H. Grass-Boada, Leandro González-Montesino, Rubén Armañanzas","doi":"10.1016/j.asoc.2025.113219","DOIUrl":"10.1016/j.asoc.2025.113219","url":null,"abstract":"<div><div>The objective of this paper is to propose a framework, called Rough Clustering-based Consensus Community Detection (RC-CCD), to effectively address the challenge of identifying community structures in complex networks from a set of different community partitions. The method uses a consensus approach based on Rough Set Theory (RST) to manage uncertainty and improve the reliability of community detection. The RC-CCD framework is tested on synthetic benchmark networks generated by the Lancichinetti–Fortunato–Radicchi (LFR) method, which simulate varying network scales, node degrees, and community sizes. Key findings demonstrate that RC-CCD outperforms established algorithms like Louvain, Greedy, and LPA in terms of normalized mutual information, showing superior accuracy and adaptability, particularly in networks with higher complexity, both in terms of size and dispersion. These results have significant implications for enhancing community detection in fields such as social and biological network analysis.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"178 ","pages":"Article 113219"},"PeriodicalIF":7.2,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144099823","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":"Advancements and prospects of fuzzy-based adaptive unscented Kalman filters for nonlinear systems: A review","authors":"Manav Kumar , Sharifuddin Mondal","doi":"10.1016/j.asoc.2025.113297","DOIUrl":"10.1016/j.asoc.2025.113297","url":null,"abstract":"<div><div>Rapid developments in computational technologies have recently imposed even more significant requirements for efficient and accurate state estimation methods that can be applied to nonlinear dynamic systems. Among the widely known nonlinear estimation techniques stands the unscented Kalman filter. However, in real-life applications, its performance is usually affected due to the presence of noise and model uncertainties. Such disturbances are handled by adaptation-based approaches, wherein the noise covariances are adjusted. Many researchers have been attracted to adaptive methods using fuzzy tuning with covariance matching in the last decade. Adaptation methodologies based on fuzzy logic applied to an unscented Kalman filter related to different practical applications are reviewed herein. It is performed by examining various kinds of fuzzy inference systems, other categories of membership functions, adaptation laws, or tuning relations of covariance matrices and their respective applications. Fuzzy logic control is one of the parts or components of artificial intelligence. The fuzzy inference systems, such as Mamdani and Takagi-Sugeno, implemented for adaptive estimation techniques with unscented Kalman filters in real-world applications are highlighted. Furthermore, the readers may easily refer to the highlighted future possibilities and significant challenges in the field.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"177 ","pages":"Article 113297"},"PeriodicalIF":7.2,"publicationDate":"2025-05-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144084375","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}
Tiejun Li , Jun Liu , Xianguo Wu , Feiming Su , Yang Liu
{"title":"Dynamic prediction and control of a tunnel boring machine with a particle swarm optimization–random forest algorithm and an integrated digital twin","authors":"Tiejun Li , Jun Liu , Xianguo Wu , Feiming Su , Yang Liu","doi":"10.1016/j.asoc.2025.113294","DOIUrl":"10.1016/j.asoc.2025.113294","url":null,"abstract":"<div><div>Tunnel boring machines (TBMs) often experience attitude deviation during excavation, impacting the stability and safety of tunnel construction. Traditional attitude adjustment relies on manual adjustment, which has a lagging effect. Therefore, the study combines a digital twin platform and a hybrid intelligence algorithm to enable real-time adjustment of the shield attitude deviation. The particle swarm optimization and random forest (PSO-RF) algorithm is first used to make accurate predictions of the shield attitude. Shapley additive explanations (SHAP) is subsequently employed to identify the key construction parameters. Then, based on these parameters, a control system for the shield attitude is designed in conjunction with a digital twin (DT) technique. A case study of China's Guiyang Metro Line 3 demonstrates the following: (1) The PSO-RF model achieves high accuracy, with R² values ranging from 0.916 to 0.943 for six shield attitude targets. (2) The key shield parameters are continuously optimized and adjusted within the control range to achieve shield attitude control. (3) The digital twin system provides real-time attitude warnings and parametric inference, significantly improving TBM performance and safety. In this paper, a novel method of combining predictive modeling and the DT platform is proposed. Under the proposed intelligent method, the attitude deviation of a TBM during tunneling was significantly reduced.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"178 ","pages":"Article 113294"},"PeriodicalIF":7.2,"publicationDate":"2025-05-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144090718","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}
Xiangyu Zhong , Fuhao Liu , Zhijiao Du , Qifeng Wan
{"title":"A two-stage consensus model incorporating manipulative and manipulated behaviors for large group decision-making under social network environment","authors":"Xiangyu Zhong , Fuhao Liu , Zhijiao Du , Qifeng Wan","doi":"10.1016/j.asoc.2025.113240","DOIUrl":"10.1016/j.asoc.2025.113240","url":null,"abstract":"<div><div>In large group decision-making (LGDM), some decision-makers (DMs) may engage in manipulative behaviors driven by personal interests, while others may become susceptible to manipulation due to the complexity and uncertainty of the decision-making process. These manipulative and manipulated behaviors hinder the effective achievement of group consensus and undermine the fairness and acceptability of the decision-making process. To address this, we propose a two-stage consensus model that accounts for both manipulative and manipulated behaviors. First, the trust relationships among DMs are adjusted based on the similarity of their evaluations, and the strength of these relationships is calculated using their adjusted mutual trust degrees. Next, a clustering method based on the fracture of relationship strength is introduced to classify DMs into subgroups. By considering DMs' hesitancy, trust relationships, and preference degrees for various alternatives expressed in their evaluations, manipulators are identified and penalized with a weight penalty. The combination of hesitation degree, trust degree, and similarities in alternative ordinals, before and after subjective adjustment, is used to identify and impose penalties on manipulated DMs. Furthermore, various objective adjustment strategies are proposed to better manage the different behaviors of DMs, thereby improving decision-making efficiency and consensus. Finally, an application example and comparative analyses are presented to validate the feasibility of the proposed method. The proposed method effectively manages manipulative and manipulated behaviors, significantly enhancing consensus efficiency, fairness, and acceptability in the decision-making process.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"178 ","pages":"Article 113240"},"PeriodicalIF":7.2,"publicationDate":"2025-05-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144134082","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}
Bama S , Hema M S , Esakkirajan S , Nageswara Guptha M
{"title":"A hierarchical transformer network with label attention for personality prediction by MBTI classification","authors":"Bama S , Hema M S , Esakkirajan S , Nageswara Guptha M","doi":"10.1016/j.asoc.2025.113267","DOIUrl":"10.1016/j.asoc.2025.113267","url":null,"abstract":"<div><div>Personality prediction is one of the emerging researches with applications in multiple fields such as psychology, ai, recommendation system, job screening, education, police department (to monitor and enforce law and order). With the evolution of deep learning models, personality prediction can be formulated as a classification problem from social media texts. More recently, transformer-based models have demonstrated impressive results in personality prediction of individuals. This paper proposes a hierarchical transformer enabled with label attention mechanism for multiclass and binary classification of personality traits from the Myers-Briggs Type Indicator (MBTI) dataset, named as Hierarchical Transformer network with Label Attention for Personality Prediction (HT-LA-PP). By capturing relationships between words in sentences using a hierarchical transformer enabled with self-attention, followed by label attention to each personality trait, the proposed model surpasses the performance of similar deep learning and Transformer-based models with significant accuracy and robustness.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"178 ","pages":"Article 113267"},"PeriodicalIF":7.2,"publicationDate":"2025-05-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144115337","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}