Applied Soft Computing最新文献

{"title":"Graph grammars and Physics Informed Neural Networks for simulating of pollution propagation on Spitzbergen","authors":"Maciej Sikora ,&nbsp;Albert Oliver-Serra ,&nbsp;Leszek Siwik ,&nbsp;Natalia Leszczyńska ,&nbsp;Tomasz Maciej Ciesielski ,&nbsp;Eirik Valseth ,&nbsp;Jacek Leszczyński ,&nbsp;Anna Paszyńska ,&nbsp;Maciej Paszyński","doi":"10.1016/j.asoc.2025.113394","DOIUrl":"10.1016/j.asoc.2025.113394","url":null,"abstract":"<div><div>In this paper, we present two computational methods for performing simulations of pollution propagation described by advection-diffusion equations. The first method employs graph grammars to describe the generation process of the computational mesh used in simulations with the meshless solver of the three-dimensional finite element method. The graph transformation rules express the three-dimensional Rivara longest-edge refinement algorithm. This solver is used for an exemplary application: performing three-dimensional simulations of pollution generation by the recently closed coal-burning power plant and the new diesel power plant, the capital of Spitzbergen. The second computational code is based on the Physics Informed Neural Networks method. It is used to calculate the dissipation of the pollution along the valley in which the city of Longyearbyen is located. We discuss the instantiation and execution of the PINN method using Google Colab implementation. There are four novelties of our paper. First, we show a lower computational cost of the proposed graph grammar model in comparison with the mesh transformations over the computational mesh. Second, we discuss the benefits and limitations of the PINN implementation of the non-stationary advection-diffusion model with respect to finite element method solvers. Third, we introduce the PINN code for non-stationary thermal inversion simulations. Fourth, using our computer simulations, we estimate the influence of the pollution from power plants on the Spitzbergen inhabitants.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"182 ","pages":"Article 113394"},"PeriodicalIF":7.2,"publicationDate":"2025-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144654163","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-criteria selection of data clustering methods for e-commerce personalization","authors":"Elżbieta Pawełek-Lubera ,&nbsp;Mateusz Przyborowski ,&nbsp;Dominik Ślęzak ,&nbsp;Adam Wasilewski","doi":"10.1016/j.asoc.2025.113559","DOIUrl":"10.1016/j.asoc.2025.113559","url":null,"abstract":"<div><div>E-commerce platforms increasingly rely on personalization to improve the user experience and drive sales, requiring efficient data clustering methods to segment users based on their behavior and preferences. However, due to the many data clustering techniques available, the key decision problem is to choose the optimal grouping method. Decision making based on single decision factors, although widely used, can lead to wrong decisions, so it is worth considering multi-criteria analysis tailored to the specifics of e-commerce customer clustering. Through extensive experiments on real e-commerce datasets, the study demonstrates the strengths and limitations of selected data clustering techniques (including the Approximated Gaussian Mixture Model, which was found to be superior to the classical Gaussian Mixture Model), considering different decision criteria related to various aspects of quality. The results provide valuable insights for e-commerce practitioners seeking to optimize their personalization strategies and ultimately suggest that a novel adaptation of the PROMETHEE II method can provide a robust framework for making informed decisions about selection of data clustering algorithms.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"182 ","pages":"Article 113559"},"PeriodicalIF":7.2,"publicationDate":"2025-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144665681","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":"Design of imitation learning fusion algorithm for mobile robotic arm control","authors":"Yang Wang ,&nbsp;Xiaoling Yan ,&nbsp;Liming Wang ,&nbsp;Wei Pan ,&nbsp;Fei Song ,&nbsp;Xinlei Zhou","doi":"10.1016/j.asoc.2025.113628","DOIUrl":"10.1016/j.asoc.2025.113628","url":null,"abstract":"<div><div>Despite the considerable potential of artificial intelligence technology in industrial applications, it still faces challenges such as high data requirements, limited generalization capabilities, and concerns with safety and stability. These issues become particularly prominent in the execution of precision operations such as opening button-lock cabinet doors during robotic inspection tasks. The tasks involve complex environmental perception, dynamically changing operational settings, high stability requirements, and the challenge of generalization, all of which necessitate the integration of multiple advanced algorithms into a fusion system design. This research focuses on the system design and algorithm integration challenges for mobile robotic arms in inspection tasks, proposing a Deep Meta-Imitation Learning (DMIL) algorithm that combines deep meta-learning with imitation learning (IL) to enhance the adaptability and efficiency of mobile robotic arms in such tasks. Expert trajectories were generated and analyzed in the CoppeliaSim simulation environment, and actual interactions were evaluated. The study employs a deep learning-based 6D pose estimation method to determine the position and orientation of button locks, with visual recognition of key operational reference points. In the imitation learning phase, the operational strategies of the robotic arm are enhanced by combining Adversarial Inverse Reinforcement Learning (AIRL) and Variable Impedance Control (VIC) technologies, supported by expert-guided trajectories and force feedback data from real-world environments. Additionally, a Latent Embedding Optimization (LEO) module is introduced into the deep meta-learning framework, enabling the model to quickly adapt to new tasks, significantly improving its generalization ability. Experiments were conducted in the CoppeliaSim simulation environment and on a mobile robotic arm platform, focusing on the recognition process, trajectory planning, and compliance control management. To assess the algorithm's effectiveness, three button-lock cabinet door-opening tasks of varying difficulty were executed. The experimental results demonstrate that the mobile robotic arm was able to accurately locate and compliantly open multiple button locks, showcasing the practicality and feasibility of this approach in advancing robotic precision operations in inspection tasks.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"182 ","pages":"Article 113628"},"PeriodicalIF":7.2,"publicationDate":"2025-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144663632","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":"Weighted classification of deep and traditional histogram-based features with kernel representation for robust facial expression recognition","authors":"Morteza Najmabadi ,&nbsp;Mina Masoudifar ,&nbsp;Ahmad Hajipour","doi":"10.1016/j.asoc.2025.113630","DOIUrl":"10.1016/j.asoc.2025.113630","url":null,"abstract":"<div><div>Feature extraction is crucial in facial expression recognition (FER) systems. This paper introduces a novel descriptor called Local Edge-based Decoded Binary Pattern (LEDB) and a lightweight 1D-CNN named Statistical Local Feature-based Network (SLFNet) to overcome limitations of deep learning approaches, such as the need for complex deep networks, high computational demands, and large training datasets. To enhance feature extraction stability, derivative-Gaussian filters are applied across four directions, yielding more robust representations. In the resulting gradient space, inter-pixel relationships are extracted to generate LEDB micropatterns, which are moderately sized yet highly discriminative, effectively capturing low-level features. Additionally, a compact 1D-CNN with 208k parameters learns high-level features from emotion-related facial regions, enhancing robustness against variations in resolution, noise, and occlusion. High-level and low-level features are fused through a weighted kernel representation strategy to increase resilience to outliers. Extensive experiments on six FER datasets—CK+ , FACES, KDEF, MMI, JAFFE, and RAF-DB—show that the proposed LEDB, SLFNet, and their combination outperform traditional handcrafted descriptors and recent deep learning techniques across various evaluation protocols. Furthermore, the system remains robust in challenging scenarios, such as those with low resolution, noise, or occlusion, which are common hurdles in FER. Code will be made available at: <span><span>https://github.com/Morteza-Najm/TDF-WKR-FER</span><svg><path></path></svg></span></div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"182 ","pages":"Article 113630"},"PeriodicalIF":7.2,"publicationDate":"2025-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144686228","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 first-order meta learning method for remaining useful life prediction of rotating machinery under limited samples","authors":"Yu Wang ,&nbsp;Shujie Liu ,&nbsp;Shuai Lv ,&nbsp;Gengshuo Liu","doi":"10.1016/j.asoc.2025.113616","DOIUrl":"10.1016/j.asoc.2025.113616","url":null,"abstract":"<div><div>Accurately predicting the remaining useful life (RUL) of rotating machinery is a challenging task in the field of Prognostics and Health Management (PHM). In practical applications, the number of samples in the target domain is often insufficient. To address this issue, we propose a First-Order Meta-Learning Network (FOMLN) to tackle the problem of equipment RUL prediction under limited samples. First, a meta-learner is constructed based on Conformer, combining the advantages of the self-attention mechanism and convolutional neural networks, enhancing the model's ability to capture both local and global features. Then, a dual-loop meta-learning strategy is designed: the inner loop learns at the sample level, modeling and updating parameters for specific tasks, while the outer loop updates the meta-parameters through task-level learning, improving the model's generalization across different tasks and its adaptability to new tasks under limited sample conditions. Extensive experimental results on the C-MAPSS dataset validate the effectiveness of the proposed method. Moreover, a practical application case study is introduced, demonstrating the model’s ability to predict the RUL of slurry pumps in an industrial site under few-shot scenarios, highlighting its potential for real-world applications.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"182 ","pages":"Article 113616"},"PeriodicalIF":7.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144702162","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An optimized convolutional neural network with a novel spherical triangular fuzzy pooling layer for an algorithmic trading model","authors":"Ehsan Mohammadian Amiri,&nbsp;Akbar Esfahanipour","doi":"10.1016/j.asoc.2025.113617","DOIUrl":"10.1016/j.asoc.2025.113617","url":null,"abstract":"<div><div>Pooling layers of Convolutional Neural Networks (CNNs) are applied to reduce the dimensionality of input features while overcoming the overfitting issue. Typically, average pooling assigns the same weight to each activation in the pooling region. However, each region in an input image may not be equally crucial in financial data. To overcome this drawback, we propose a novel pooling layer that incorporates spherical fuzzy logic to account for the inherent uncertainty of financial data in a novel algorithmic trading model. To that end, we propose two minor models for prioritizing activations based on technical indicators of extreme points and time positions, respectively. Based on these two models, we propose an Indicator-Time-based Spherical Triangular Fuzzy pooling (ITSFP) model to capture critical market signals with greater accuracy and adapt more effectively to market conditions. To optimize the algorithmic trading model, a Genetic Algorithm (GA) has been designed to fine-tune the architecture of the proposed CNN, improving its adaptability and performance. The results demonstrate that the optimized ITSFP outperformed the others, achieving an accuracy of 80.08 %, while the optimized Fuzzy Pooling (FP), Max Pooling (MP), and Average Pooling (AP) achieved accuracies of 69.84 %, 63.60 %, and 59.77 %, respectively. The algorithmic trading based on the ITSFP model achieved the highest compound return (2.5866), Sharpe ratio (0.3632), and Sortino ratio (0.5242), reflecting its superior risk-adjusted returns and recorded the lowest Maximum Drawdown (2.8632), indicating superior resilience during market downturns. The Wilcoxon signed-rank test has been applied to show significant outperformance of the proposed ITSFP against others.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"182 ","pages":"Article 113617"},"PeriodicalIF":7.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144685482","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 learning-based feature fusion and Forecasting approach for stock market Prediction","authors":"Tzu-Chia Chen","doi":"10.1016/j.asoc.2025.113623","DOIUrl":"10.1016/j.asoc.2025.113623","url":null,"abstract":"<div><div>Generally, stock market prediction is a current renowned research topic. The existing prediction approaches are on the basis of the econometric and statistical approaches. Nevertheless, these approaches are complex to pact with non-stationary time series data. Thus, this study develops a new approach for stock market prediction utilizing a hybrid deep learning approach. In this work, the pre-processing stage is done initially with the assistance of yeo-jhonson transformation and padding-based Fourier transform (FT) denoising model. After that, technical indicators, like Williams’s %R, Rate of Change, Triple Exponential Moving Average (TRIX), Average Directional Index (ADX), Average True Range (ATR), and Relative Strength Index (RSI), are extracted. Subsequently, the feature fusion procedure is done by utilizing Morisita's overlap index and Deep Belief Network (DBN) model. Lastly, stock market forecasting is done by a hybrid approach integrating Deep Long Short-Term Memory (Deep LSTM) and Multi-Layer Perceptron (MLP). Moreover, the proposed model is compared with the conventional models, such as Padding-based Fourier Transform Denoising (P-FTD)+Recurrent Neural Network (RNN), Feedforward Neural Network (FNN)+Back-propagation Neural Network (BPNN), Residual-CNN-Seq2Seq (RCSNet), Long short-term memory (LSTM), Competitive Feedback Particle Swarm Optimization-based Deep Recurrent Neural Network (CFPSO-based Deep RNN) and Rider Deep LSTM &amp; Deep RNN. Finally, the experimentation analysis states that the performance of Deep LSTM-MLP is superior to conventional approaches regarding the MSE, RMSE and MAE with the values of 0.113, 0.337, and 0.169.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"182 ","pages":"Article 113623"},"PeriodicalIF":7.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144680392","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":"Cloud detection network based on context feature enhancement for remote sensing images","authors":"Baotong Su ,&nbsp;Yao Chen ,&nbsp;Wenguang Zheng","doi":"10.1016/j.asoc.2025.113553","DOIUrl":"10.1016/j.asoc.2025.113553","url":null,"abstract":"<div><div>Cloud detection aims to identify cloud regions in satellite remote sensing images. This task remains particularly challenging due to the diverse morphological characteristics of clouds and their spectral similarity to bright ground surfaces, such as snow or sand. Most existing methods still suffer from unsatisfactory performance, primarily due to their inability to effectively model global information, which restricts the model’s understanding of the differences between cloud regions and the background. To address these issues, we propose a novel framework called the Contextual Feature Enhancement Network (CFEnet), which enhances contextual representations by incorporating global and multi-scale information. Specifically, we first introduce the global information modeling module (GIMM), which captures long-range dependencies and global context at multiple scales, enabling the model to comprehensively perceive image features. Subsequently, we design a multi-scale feature pyramid (MSFP) to gradually integrate high-resolution low-level features with low-resolution high-level features. Finally, a feature fusion module (FFM) is employed to fuse feature maps from different modules. Our method’s effectiveness was evaluated on three public datasets: GF-1 WFV, LandSat8, and SPARCS. Experimental results demonstrated that CFEnet significantly outperforms several state-of-the-art methods, achieving an average improvement of 1.77% in detection precision.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"182 ","pages":"Article 113553"},"PeriodicalIF":7.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144671097","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 dual-layer state space network for Multivariate Time Series Forecasting with dimensional interdependency","authors":"Ning Zhou,&nbsp;Linfu Sun","doi":"10.1016/j.asoc.2025.113452","DOIUrl":"10.1016/j.asoc.2025.113452","url":null,"abstract":"<div><div>Multivariate Time Series Forecasting (MTSF) presents significant challenges due to the need to capture intricate inter-dimensional dependencies and temporal dynamics. Existing Transformer-based models often suffer from quadratic computational complexity and limited efficacy in modeling long-range temporal dependencies. To overcome these limitations, we propose Mixed Spatio-Temporal Mamba (MST-Mamba), a novel architecture that integrates mixed-dimensional dependency encoding, sliding-window patch processing, and hierarchical state-space modeling. Specifically, a spectral correlation-based mixed-dimensional encoder that effectively captures cross-dimensional dependencies by analyzing amplitude spectrum similarities. Intra-patch temporal features are modeled using a Bidirectional Structured State Space architecture, enabling efficient representation of local temporal patterns. For inter-patch temporal modeling, the Mamba module is employed, facilitating robust and scalable long-range dependency extraction. We evaluate the MST-Mamba architecture on several real-world datasets, Experimental results demonstrate that our model achieves an improvement in AVG MSE of 12.9%, 4.6%, 3.1%, and 3.3% respectively across PEMS07, Traffic, Weather and Electricity datasets. Moreover, our model demonstrates superior performance in handling long sequences with lookback information and exhibits enhanced robustness in noisy environments.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"183 ","pages":"Article 113452"},"PeriodicalIF":7.2,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144703048","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 comparative trade-off analysis on accuracy and efficiency for federated learning in demand forecasting","authors":"Hang Qi,&nbsp;Jieping Luo,&nbsp;Qiyue Li,&nbsp;Jingjin Wu","doi":"10.1016/j.asoc.2025.113561","DOIUrl":"10.1016/j.asoc.2025.113561","url":null,"abstract":"<div><div>Federated Learning (FL) is an emerging learning mechanism that can achieve accuracy, efficiency, and privacy concurrently, which could be particularly useful in scenarios where large volumes of sensitive data are involved, such as demand forecasting in inventory management. In this paper, we consider a dataset composing of sales records of multiple products across ten different Walmart stores in the USA, and conduct a comparative study of centralized learning, distributed learning, and FL, focusing on the accuracy of predicting future demand of certain products and the required volume of data for transmission by the multi-layer perceptron model. Our results demonstrate that, with the same number of training rounds, FL achieves competitive accuracy in predicting future demands across the selected product categories while significantly reducing data transmission compared to other learning approaches, highlighting the efficiency and practicality of FL. In addition, we compare the performances of FL approaches with different combinations of store-level data from various regions, and examine the trade-off analysis in terms of accuracy and transmission efficiency under different scenarios.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"182 ","pages":"Article 113561"},"PeriodicalIF":7.2,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144654307","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}