Applied Soft Computing最新文献

{"title":"A multi-granularity decision tree algorithm based on variable precision rough sets and Zentropy","authors":"Hui Dong ,&nbsp;Caihui Liu ,&nbsp;Xiying Chen ,&nbsp;Duoqian Miao","doi":"10.1016/j.asoc.2025.113851","DOIUrl":"10.1016/j.asoc.2025.113851","url":null,"abstract":"<div><div>The existing decision tree algorithms often use a single-layer measure to process data, which cannot fully consider the complex interactions and dependencies between different granularity levels. In addition, decision tree algorithms inevitably face the issue of multi-value preference, which may lead to the selection of unreasonable attributes in the process of partition, thereby affecting the performance of the algorithms. Therefore, this paper proposes an improved decision tree algorithm, called Ze-VNDT, which combines variable precision rough sets with Zentropy. First, to avoid the information loss caused by data discretization, this paper introduces variable precision neighborhood rough sets for data processing. Second, by analyzing the granularity level structure within the variable precision neighborhood rough set model, knowledge uncertainty is analyzed from three granularity levels: decision classes, approximate relations, and similarity classes. We describe the uncertain knowledge from the overall to the internal using the idea of going from coarse to fine, and design a Zentropy to measure uncertainty. To address the issue of multi-value preference, an adaptive weighted Zentropy uncertainty measure is designed based on the definition of uncertainty measure based on Zentropy. Third, when constructing the improved decision tree algorithm, the optimal attributes are selected based on the designed uncertainty measure. Finally, numerical experiments on 18 UCI datasets validated the effectiveness and rationality of the proposed algorithm. The experimental results showed that, compared to traditional algorithms and the latest improved algorithms, the proposed algorithm achieved an average accuracy of 94.79%, an average precision of 85.77%, an average recall rate of 84.68%, and an F1-score of 84.97% across the 18 datasets. It ranked first in all five evaluation metrics, demonstrating higher stability and accuracy.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"185 ","pages":"Article 113851"},"PeriodicalIF":6.6,"publicationDate":"2025-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145050477","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":"Mitigating over-squashing in graph few-shot learning by leveraging local and global similarities","authors":"Yassin Mohamadi,&nbsp;Mostafa Haghir Chehreghani","doi":"10.1016/j.asoc.2025.113863","DOIUrl":"10.1016/j.asoc.2025.113863","url":null,"abstract":"<div><div>Supervised machine learning models, particularly neural networks, often fail to deliver satisfactory results in scenarios with insufficient data. This becomes even more challenging when dealing with inherently complex data, such as graph data. This paper addresses the issue of learning with a limited number of samples, known as <span><math><mi>n</mi></math></span>-way <span><math><mi>k</mi></math></span>-shot learning, within the context of graph data. Our research extends the concept of similarity from neighboring nodes to the entire graph by leveraging transitivity relations. By employing edges and strong transitivity relations, we utilize a bipartite graph neural network that capitalizes on both local neighborhoods and distant, yet similar, nodes to generate node embeddings. This approach has demonstrated effectiveness in tasks such as node classification. Our proposed model’s ability to mitigate the over-squashing problem enhances its generalizability, resulting in a task-invariant model. Experimental results on various graph datasets show that the embeddings produced by our model are not task-specific. Consequently, our model outperforms other models in few-shot learning scenarios, where only a limited number of labeled nodes are available for each distinct downstream task.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"184 ","pages":"Article 113863"},"PeriodicalIF":6.6,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145026539","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":"CMFF: A cross-modal multi-layer feature fusion network for multimodal sentiment analysis","authors":"Shuting Zheng ,&nbsp;Jingling Zhang ,&nbsp;Yuanzhao Deng ,&nbsp;Lanxiang Chen","doi":"10.1016/j.asoc.2025.113868","DOIUrl":"10.1016/j.asoc.2025.113868","url":null,"abstract":"<div><div>Multimodal sentiment analysis seeks to interpret speaker sentiment by integrating information from multiple modalities, typically text and audio. While existing methods often focus on fusing deep-layer features extracted from the final stages of unimodal encoders, they may overlook crucial fine-grained information present in shallow-layer features (e.g., subtle phonetic variations or basic syntactic structures) relevant for nuanced sentiment understanding. Furthermore, effectively fusing features from different modalities presents the dual challenges of dynamically weighting each modality’s contribution and accommodating their inherent data heterogeneity. To address these limitations, we propose a novel Cross-modal Multi-layer Feature Fusion (CMFF) network. CMFF explicitly leverages the hierarchical information contained in both shallow-layer and deep-layer features from text and audio modalities. It employs multi-head cross-modal attention mechanisms within its fusion layers to facilitate interaction across feature layers and modalities. Crucially, CMFF incorporates a Mixture of Gated Experts (MoGE) network within these fusion layers. The MoGE utilizes modality-specific expert sub-networks, each tailored to process the distinct characteristics of text or audio data, thereby directly addressing data heterogeneity. Concurrently, each expert employs an internal gated feed-forward mechanism. This allows the model to dynamically control the information flow for each feature vector, effectively learning to weigh the importance of different feature dimensions from each layer and modality based on the input context. Extensive experiments conducted on the benchmark CMU-MOSI and CMU-MOSEI datasets demonstrate that the proposed CMFF model achieves competitive or superior performance compared to state-of-the-art methods across various standard evaluation metrics.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"184 ","pages":"Article 113868"},"PeriodicalIF":6.6,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145048810","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":"AOT-PixelNet: Lightweight and interpretable detection of forged images via adaptive orthogonal transform","authors":"Dengtai Tan ,&nbsp;Deyi Yang ,&nbsp;Boao Tan ,&nbsp;Chengyu Niu ,&nbsp;Yang Yang ,&nbsp;Shichao Li","doi":"10.1016/j.asoc.2025.113873","DOIUrl":"10.1016/j.asoc.2025.113873","url":null,"abstract":"<div><div>Generative image detection faces persistent challenges in terms of generalization and interpretability, limiting its reliability in complex scenarios. To address these issues, we propose AOT-PixelNet, a lightweight and interpretable detection framework that integrates an Adaptive Orthogonal Transform (AOT) module with a streamlined 1 × 1 convolution-based PixelNet architecture. The AOT module leverages diverse orthogonal transforms, such as FFT and DCT, to extract informative frequency-domain features, thereby enhancing sensitivity to medium- and high-frequency artifacts. Meanwhile, PixelNet minimizes parameter count (only 0.98 million) while effectively capturing cross-channel inconsistencies and mitigating overfitting. Experimental evaluations on multiple unseen GAN and diffusion-based datasets demonstrate that AOT-PixelNet achieves superior performance with minimal computational cost. Specifically, it outperforms the NPR method by 0.6% and 11.76% on the ForenSynths and GenImage datasets, respectively, validating the framework’s robustness, effectiveness, and interpretability.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"185 ","pages":"Article 113873"},"PeriodicalIF":6.6,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145050473","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":"TPLLM: A traffic prediction framework based on pretrained Large Language Models","authors":"Tian Ma ,&nbsp;Yixuan Zhao ,&nbsp;Minda Li ,&nbsp;Yue Chen ,&nbsp;Fangshu Lei ,&nbsp;Yanan Zhao ,&nbsp;Maazen Alsabaan","doi":"10.1016/j.asoc.2025.113840","DOIUrl":"10.1016/j.asoc.2025.113840","url":null,"abstract":"<div><div>Traffic prediction constitutes a critical component in sustainable urban data analysis, playing a pivotal role in optimizing transportation systems for reduced carbon emissions and improved energy efficiency. The precision of prevailing deep learning-driven traffic prediction models typically improves as the volume of training data increases. However, the procurement of comprehensive spatiotemporal datasets for traffic is often fraught with challenges, primarily stemming from the substantial costs associated with data collection and retention. This limitation severely hinders the deployment of models in regions with insufficient historical data. Consequently, developing a model that can achieve accurate predictions and good generalization ability in areas with limited historical traffic data is a challenging problem. It is noteworthy that the rapidly advancing pretrained Large Language Models (LLMs) of recent years demonstrate exceptional proficiency in cross-modality knowledge transfer and few-shot learning. Recognizing the sequential nature of traffic data, similar to language, we introduce TPLLM, a novel traffic prediction framework leveraging LLMs. In this framework, we construct a sequence embedding layer based on Convolutional Neural Networks (CNNs) and a graph embedding layer based on Graph Convolutional Networks (GCNs) to extract sequence features and spatial features, respectively. These are subsequently integrated to form inputs that are suitable for LLMs. A Low-Rank Adaptation (LoRA) fine-tuning approach is applied to TPLLM, thereby facilitating efficient learning and minimizing computational demands. Experiments on two real-world datasets demonstrate that TPLLM exhibits commendable performance in both full-sample and few-shot prediction scenarios.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"184 ","pages":"Article 113840"},"PeriodicalIF":6.6,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145049465","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":"Quantum fuzzy logic for edge detection: A demonstration on NISQ hardware","authors":"G. Nunziata ,&nbsp;S. Crisci ,&nbsp;G. De Gregorio ,&nbsp;R. Schiattarella ,&nbsp;G. Acampora ,&nbsp;L. Coraggio ,&nbsp;N. Itaco","doi":"10.1016/j.asoc.2025.113866","DOIUrl":"10.1016/j.asoc.2025.113866","url":null,"abstract":"<div><div>Quantum computing offers the potential to enhance computational efficiency beyond classical methods, but practical implementation remains challenging due to the limitations of Noisy Intermediate-Scale Quantum (NISQ) hardware, namely, restricted qubit counts, limited connectivity, and the presence of noise and decoherence. This study presents a novel approach to edge detection by leveraging a recently developed Quantum Fuzzy Inference Engine, implemented on a NISQ device. We introduce an optimized quantum circuit for its implementation, reducing qubit requirements and gate depth to improve execution on NISQ hardware. To overcome constraints related to large-scale image processing, a hybrid quantum–classical lookup table approach is employed. Edge detection performance is evaluated on the Berkeley Segmentation Data Set and Benchmarks 500 dataset under different conditions, including classical execution, ideal quantum simulation, noisy quantum simulation, and NISQ hardware calculation. Results demonstrate that the quantum fuzzy logic-based edge detection achieves outcomes comparable to classical methods by using fewer operations, marking a step toward practical quantum-enhanced image processing.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"185 ","pages":"Article 113866"},"PeriodicalIF":6.6,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145050479","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":"Dual-splitting conformal prediction for multi-step time series forecasting","authors":"Qingdi Yu ,&nbsp;Zhiwei Cao ,&nbsp;Ruihang Wang ,&nbsp;Zhen Yang ,&nbsp;Lijun Deng ,&nbsp;Min Hu ,&nbsp;Yong Luo ,&nbsp;Xin Zhou","doi":"10.1016/j.asoc.2025.113825","DOIUrl":"10.1016/j.asoc.2025.113825","url":null,"abstract":"<div><div>Time series forecasting is crucial for applications like resource scheduling and risk management, where multi-step predictions provide a comprehensive view of future trends. Uncertainty Quantification (UQ) is a mainstream approach for addressing forecasting uncertainties, with Conformal Prediction (CP) gaining attention due to its model-agnostic nature and statistical guarantees. However, most variants of CP are designed for single-step predictions and face challenges in multi-step scenarios, such as reliance on real-time data and limited scalability. This highlights the need for CP methods specifically tailored to multi-step forecasting. We propose the Dual-Splitting Conformal Prediction (DSCP) method, a novel CP approach designed to capture inherent dependencies within time-series data for <strong>multi-step</strong> forecasting. Experimental results on real-world datasets from four different domains demonstrate that DSCP significantly outperforms existing CP variants in terms of the Winkler Score, improving performance by up to 23.59% compared to state-of-the-art methods. Furthermore, the deployment of DSCP for renewable energy generation and IT load forecasting in the power management of a real-world trajectory-based application achieves an 11.25% reduction in carbon emissions through predictive optimization of data center operations and control strategies.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"184 ","pages":"Article 113825"},"PeriodicalIF":6.6,"publicationDate":"2025-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145026534","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":"FDP-Net: Fourier transform guided lightweight depthwise and pointwise dynamic pooling based neural network for medical image classification","authors":"Asfak Ali ,&nbsp;Rajdeep Pal ,&nbsp;Aishik Paul ,&nbsp;Ram Sarkar","doi":"10.1016/j.asoc.2025.113824","DOIUrl":"10.1016/j.asoc.2025.113824","url":null,"abstract":"<div><div>Nowadays, deep learning-based medical image classification has become essential, especially in developing countries because of the high volume of patients with less medical professionals as well as required infrastructures. Deep learning models often help in the early detection of diseases; however, it require a high amount of processing power, and sometimes it becomes less scalable for various computer-aided diagnosis. To this end, in this paper, a lightweight Fourier Transform guided Depth and Pointwise Dynamic Pooling based Neural Network (FDP-Net), has been proposed for medical image classification. This paper introduces a Depth and Pointwise Feature Fusion (DPFF) block for learning the important features with less computation and without increasing the model parameters. It also proposes a dynamic pooling technique, an alternative to traditional max-pooling, which dynamically selects the important features. The proposed FDP-Net model is trained to classify medical images with the guidance of Fourier Transformation and multitask loss function, which makes the model converge faster and reduces overfitting. The proposed model has been tested on Acute Lymphoblastic Leukemia (ALL) dataset, Peripheral Blood Cell (PBC) dataset, and Raabin White blood Cell (Raabin-WBC) dataset, and it outperforms the state-of-the-art models with 100%, 98.13% and 96.79% classification accuracies, respectively. Additionally, the proposed model is made with only 0.349 million parameters, thereby enabling faster processing. Code will be avilabe at <span><span>https://github.com/asfakali/FDP-Net</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"184 ","pages":"Article 113824"},"PeriodicalIF":6.6,"publicationDate":"2025-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145026537","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":"Advanced YOLOv4 for real-time underwater object detection: An application-oriented approach","authors":"Pratima Sarkar ,&nbsp;Sourav De ,&nbsp;Prasenjit Dey ,&nbsp;Sandeep Gurung","doi":"10.1016/j.asoc.2025.113837","DOIUrl":"10.1016/j.asoc.2025.113837","url":null,"abstract":"<div><div>Detecting debris and monitoring marine life in sea aquaculture face challenges due to limited visibility and the presence of diverse. Underwater object detection by Autonomous Unmanned Vehicle(AUV) is inherently more challenging than land due to light attenuation and water turbidity, especially for small and dense objects in murky images, where extracting high-quality features is hindered. In this paper, we present an efficient approach for real-time underwater object detection through improvements in image enhancement, data augmentation, and feature aggregation. Initially, U-Shape Transformer is applied to enhance the original images. For data augmentation, it is observable that while Mosaic data augmentation enhances complex images but fails to improve small-object detection due generation of less number of images with small objects. To address this limitation, we propose Underwater-Mosaic (U-Mosaic), a modified Mosaic data augmentation technique designed to enhance small-object detection. Additionally, it was noted that existing YOLOv4 struggles with detecting small and densely populated objects in underwater images as unable to get sufficient features for small objects due to downsampling, image quality and also found difficulty in selecting anchor box size. Therefore, we propose a model called Advanced YOLOv4, tailored for underwater object detection. The proposed Advanced YOLOv4 aims to improve object detection efficiency by altering the neck and prediction layers of YOLOv4. Moreover, we introduce an additional spatial pyramid pooling layer to aggregate features and reduce feature dimensions thereby improving object detection rates. Also, the proposed work concentrates on very large object detection and for this purpose used downsampling during the detection of large objects. The proposed approach is validated through two distinct application areas: (i) detecting and locating debris (ii) detecting fish from underwater images. For validation, the Trash ICRA19 dataset is used for debris detection, while the Brackish dataset is employed for fish detection. UIQM and UCIQE, image enhancement assessment metrics are used to measure quality of enhanced images and found more than 20% better result for both the datasets. The proposed real-time underwater object detection model outperformed single-stage object detectors like YOLOv3, YOLOv4, YOLOv5, YOLOv7, and KPE-YOLOv5 by 5% in terms of mean Average Precision(mAP). Also proposed work compared with two-stage detector RCNN and found 8% better mAP than RCNN.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"185 ","pages":"Article 113837"},"PeriodicalIF":6.6,"publicationDate":"2025-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145050474","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":"Skin lesion classification with mini-batch sampling and deep metric learning","authors":"Shengdan Hu ,&nbsp;Zhifei Zhang ,&nbsp;Li Ying ,&nbsp;Guangming Lang","doi":"10.1016/j.asoc.2025.113850","DOIUrl":"10.1016/j.asoc.2025.113850","url":null,"abstract":"<div><div>Skin lesion image classification based on deep learning has recently garnered significant attention. However, directly applying methods that perform well in general computer vision tasks to skin lesion image classification is not ideal, as skin lesion image datasets possess intrinsic characteristics, such as class imbalance, intra-class variability, and inter-class similarity. To tackle these challenges simultaneously, we propose a novel unified learning framework, named mBSML, which integrates mini-batch sampling and deep metric learning. In this framework, mini-batch sampling re-samples data in real-time during each iteration of learning, while a new loss function combines mini-batch distance metric-based loss with cross-entropy loss. Through the alternating training procedure on both imbalanced training data and balanced re-sampling data, mBSML effectively learns from global distribution information and local similarity information, not only from the original dataset but also from the minority classes. Extensive experiments conducted on two publicly available datasets demonstrate the effectiveness of mBSML for skin lesion image classification.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"185 ","pages":"Article 113850"},"PeriodicalIF":6.6,"publicationDate":"2025-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145050481","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}