Applied Intelligence最新文献_第5页

EPDNet: Light-weight small target detection algorithm based on pruning and logical distillation EPDNet：基于剪枝和逻辑蒸馏的轻量级小目标检测算法

IF 3.4 2区计算机科学

Applied Intelligence Pub Date : 2025-04-28 DOI: 10.1007/s10489-025-06582-3

Gaofeng Zhu, Zhixue Wang, Fenghua Zhu, Gang Xiong, Zheng Li

{"title":"EPDNet: Light-weight small target detection algorithm based on pruning and logical distillation","authors":"Gaofeng Zhu, Zhixue Wang, Fenghua Zhu, Gang Xiong, Zheng Li","doi":"10.1007/s10489-025-06582-3","DOIUrl":"10.1007/s10489-025-06582-3","url":null,"abstract":"<div><p>Drone detection technology plays a crucial role in various fields. However, due to the limited computational resources of edge devices onboard drones, achieving efficient detection using large-parameter algorithms remains challenging. Small target detection in drone-based applications faces several difficulties, including the small size of targets, limited feature information, and vulnerability to environmental interference. Moreover, existing lightweight small target detection methods often compromise detection accuracy while reducing model parameters, failing to meet the dual requirements of accuracy and efficiency in drone scenarios. To address these challenges, this paper proposes EPDNet, a lightweight small target detection algorithm designed for drone applications. First, ConvNextV2 replaces the original backbone network, incorporating a fully convolutional masked autoencoder framework combined with a self-supervised learning strategy to enhance the extraction of essential low-level features. Additionally, the EC2f feature extraction module is introduced to enable interactive modeling of contextual detail features across different target scales, orientations, and shapes. Furthermore, an adaptive channel pruning scheme is designed to reduce redundant parameters and computational complexity, thereby enhancing algorithm efficiency. Finally, the detection performance of the pruned model is further optimized using knowledge distillation. Experimental results on the VisDrone2019 aerial photography dataset demonstrate that EPDNet improves detection precision (P) by 2.6%, increases mean average precision (mAP) by 3.0%, reduces the number of parameters by 29.6%, and decreases computational cost by 17.8%. These results indicate that EPDNet effectively meets the lightweight deployment requirements of drone-based applications.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143879639","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Orthogonal and spherical quaternion features for weakly supervised learning with label confidence optimization 带标签置信度优化的弱监督学习的正交和球面四元数特征

IF 3.4 2区计算机科学

Applied Intelligence Pub Date : 2025-04-28 DOI: 10.1007/s10489-025-06575-2

Heng Zhou, Ping Zhong

{"title":"Orthogonal and spherical quaternion features for weakly supervised learning with label confidence optimization","authors":"Heng Zhou, Ping Zhong","doi":"10.1007/s10489-025-06575-2","DOIUrl":"10.1007/s10489-025-06575-2","url":null,"abstract":"<div><p>Weakly supervised learning (WSL) addresses the challenge of incomplete or noisy labels, but current methods often fail to capture the complexities introduced by weak labels in feature extraction, revealing the limitations of neural networks in modeling the intricate relationships between features and labels. To address these issues, we introduce the Orthogonal and Spherical Quaternion Neural Network (OSQNN), which utilizes quaternion feature embedding with an orthogonal constraint to map real-valued features into quaternion space. This approach improves the understanding of feature-label relationships by overcoming the challenge of embedding real-world data into quaternion spaces. OSQNN maps quaternion features onto a sphere and estimates label reliability through nearest neighbors, maintaining a coherent geometric structure in feature distributions. Furthermore, quaternion convolutions are transformed into parallel grouped real-valued convolutions, enhancing processing efficiency without sacrificing the benefits of quaternion-based computations. Additionally, we propose the Label Confidence Guided Expectation-Maximization (LCGEM) algorithm, integrated into OSQNN, to more effectively capture the complex relationships between weak labels and feature distributions. Experimental results across eight datasets demonstrate the superiority of OSQNN. For instance, in SSL on CIFAR10 (20% labeled data) and CIFAR100, it achieved 91.06% and 69.16% accuracy respectively; in NSL with 40% incorrect labels on CIFAR10 and CIFAR100, the accuracies were 80.84% and 51.98%, showing its high accuracy and robustness. The ablation study highlights the role of the orthogonal constraint and spherical feature mapping in improving performance, while t-SNE visualization confirms the ability of OSQNN to learn discriminative feature representations.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-04-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143879695","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

ATTD and ATDS detecting abnormal trajectory detection for urban traffic data atd和ATDS检测对城市交通数据的异常轨迹检测

IF 3.4 2区计算机科学

Applied Intelligence Pub Date : 2025-04-26 DOI: 10.1007/s10489-025-06370-z

Xi-Te Wang, Zheng Xu, Xiao-Yue Liao, Mei Bai, Qian Ma

{"title":"ATTD and ATDS detecting abnormal trajectory detection for urban traffic data","authors":"Xi-Te Wang, Zheng Xu, Xiao-Yue Liao, Mei Bai, Qian Ma","doi":"10.1007/s10489-025-06370-z","DOIUrl":"10.1007/s10489-025-06370-z","url":null,"abstract":"<div><p>Abnormal trajectory detection is pivotal for ensuring safety and optimizing operations in urban traffic management. Despite the progress in this field, current anomaly detection methods, such as the Spatial-Temporal Relationship (STR) algorithm, face limitations including high computational complexity due to simultaneous model calculations, delayed anomaly detection, and an inability to estimate anomalies in the remaining route during online detection. These limitations can lead to inefficiencies and reduced safety in real-world applications. In this paper, we address these limitations by introducing two novel algorithms: Anomaly Trajectory Detection based on Temporal model (ATTD) and Abnormal Trajectory Detection based on Dual Standards (ATDS). The ATTD algorithm simplifies the detection process by integrating a unified spatio-temporal model, which reduces computational complexity and accelerates the detection of anomalies. Furthermore, the ATDS algorithm introduces a proactive approach to anomaly detection that not only identifies anomalies in real-time but also predicts potential deviations in the remaining trajectory, thus providing a more comprehensive and timely detection mechanism. Through extensive experiments on real taxi trajectory datasets, we demonstrate that our algorithms significantly outperform the STR algorithm and other existing methods in terms of detection accuracy and computational efficiency. Our work contributes to the field by providing a more robust and efficient approach to anomaly trajectory detection.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-04-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143875405","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

HLNet: high-level attention mechanism U-Net + + for brain tumor segmentation in MRI 高水平注意机制U-Net + +在MRI脑肿瘤分割中的应用

IF 3.4 2区计算机科学

Applied Intelligence Pub Date : 2025-04-25 DOI: 10.1007/s10489-025-06568-1

Wenyang Yang, Zhiming Li, Chao Du, Steven Kwok Keung Chow

{"title":"HLNet: high-level attention mechanism U-Net + + for brain tumor segmentation in MRI","authors":"Wenyang Yang, Zhiming Li, Chao Du, Steven Kwok Keung Chow","doi":"10.1007/s10489-025-06568-1","DOIUrl":"10.1007/s10489-025-06568-1","url":null,"abstract":"<div><p>The high-level attention mechanism enhances object detection by focusing on important features and details, making it a potential tool for tumor segmentation. However, its effectiveness and efficiency in this context remain uncertain. This study aims to investigate the efficiency, feasibility and effectiveness of integrating a high-level attention mechanism into the U-Net and U-Net + + model for improving tumor segmentation. Experiments were conducted using U-Net and U-Net + + models augmented with high-level attention mechanisms to compare their performance. The proposed model incorporated high-level attention mechanisms in the encoder, decoder, and skip connections. Model training and validation were performed using T1, FLAIR, T2, and T1ce MR images from the BraTS2018 and BraTS2019 datasets. To further evaluate the model's effectiveness, testing was conducted on the UPenn-GBM dataset provided by the Center for Biomedical Image Computing and Analysis at the University of Pennsylvania. The segmentation accuracy of the high-level attention U-Net + + was evaluated using the DICE score, achieving values of 88.68 (ET), 89.71 (TC), and 91.50 (WT) on the BraTS2019 dataset and 90.93 (ET), 92.79 (TC), and 93.77 (WT) on the UPEEN-GBM dataset. The results demonstrate that U-Net + + integrated with the high-level attention mechanism achieves higher accuracy in brain tumor segmentation compared to baseline models. Experiments conducted on comparable and challenging datasets highlight the superior performance of the proposed approach. Furthermore, the proposed model exhibits promising potential for generalization to other datasets or use cases, making it a viable tool for broader medical imaging applications.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-04-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10489-025-06568-1.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143871330","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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}

引用次数: 0

Multimodal intent recognition based on text-guided cross-modal attention 基于文本引导的跨模态注意力的多模态意图识别

IF 3.4 2区计算机科学

Applied Intelligence Pub Date : 2025-04-25 DOI: 10.1007/s10489-025-06583-2

Zhengyi Li, Junjie Peng, Xuanchao Lin, Zesu Cai

{"title":"Multimodal intent recognition based on text-guided cross-modal attention","authors":"Zhengyi Li, Junjie Peng, Xuanchao Lin, Zesu Cai","doi":"10.1007/s10489-025-06583-2","DOIUrl":"10.1007/s10489-025-06583-2","url":null,"abstract":"<div><p>In natural language understanding, intent recognition stands out as a crucial task that has drawn significant attention. While previous research focuses on intent recognition using task-specific unimodal data, real-world scenarios often involve human intents expressed through various ways, including speech, tone of voice, facial expressions, and actions. This prompts research into integrating multimodal information to more accurately identify human intent. However, existing intent recognition studies often fuse textual and non-textual modalities without considering their quality gap. The gap in feature quality across different modalities hinders the improvement of the model’s performance. To address this challenge, we propose a multimodal intent recognition model to enhance non-textual modality features. Specifically, we enrich the semantics of non-textual modalities by replacing redundant information through text-guided cross-modal attention. Additionally, we introduce a text-centric adaptive fusion gating mechanism to capitalize on the primary role of text modality in intent recognition. Extensive experiments on two multimodal task datasets show that our proposed model performs better in all metrics than state-of-the-art multimodal models. The results demonstrate that our model efficiently enhances non-textual modality features and fuses multimodal information, showing promising potential for intent recognition.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-04-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143871333","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Breaking barriers in hotspot mining: a novel approach to reflecting domain characteristics and correlations 打破热点挖掘的障碍：反映领域特征和相关性的新方法

IF 3.4 2区计算机科学

Applied Intelligence Pub Date : 2025-04-25 DOI: 10.1007/s10489-024-06136-z

Wei Chen, Zhengtao Yu, Shengxiang Gao, Yantuan Xian

{"title":"Breaking barriers in hotspot mining: a novel approach to reflecting domain characteristics and correlations","authors":"Wei Chen, Zhengtao Yu, Shengxiang Gao, Yantuan Xian","doi":"10.1007/s10489-024-06136-z","DOIUrl":"10.1007/s10489-024-06136-z","url":null,"abstract":"<div><p>Hotspot mining is essential for acquiring information on hotspots and knowledge in a given domain, and it is also of great value for improving the efficiency and quality of scientific research work in the profession. Previous literature on hotspot mining did not take into account the domain characteristics of the literature and the diverse associations of the domain-specific literature itself. It is a challenging task to reflect the domain characteristics of the literature and use multiple correlations among the literature in the model. In this study, we depict each association link using a heterogeneous network of metallurgical literature and simultaneously fuse metallurgical domain-specific knowledge by aggregating the knowledge graph data of the neighbors into the term nodes of the heterogeneous network of the literature. A proposed heterogeneous academic network metallurgical literature hotspot mining method incorporates domain-specific knowledge. This method reflects various types of associational relation information in the literature via the heterogeneous network. In the meantime, it weights and analyzes the paths in the heterogeneous network, identifies the most critical paths for vectorized representation, and highlights the impact of essential paths and domain knowledge on representation learning, enhancing the information representation of diverse data in the model and improving its accuracy. The suggested model is compared with GCN, the MAGNN standard model, and its ablation model as applied to public and metallurgical literature datasets. The findings on the public dataset show that the proposed method is superior to the other two approaches. In contrast, the results for the metallurgical literature dataset are more conspicuous, with the proposed method exhibiting a more remarkable improvement in HR and NGCC.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-04-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143871332","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An adaptive quantitative trading strategy optimization framework based on meta reinforcement learning and cognitive game theory 基于元强化学习和认知博弈论的自适应定量交易策略优化框架

IF 3.4 2区计算机科学

Applied Intelligence Pub Date : 2025-04-25 DOI: 10.1007/s10489-025-06423-3

Zhiheng Shen, Hanchi Huang

{"title":"An adaptive quantitative trading strategy optimization framework based on meta reinforcement learning and cognitive game theory","authors":"Zhiheng Shen, Hanchi Huang","doi":"10.1007/s10489-025-06423-3","DOIUrl":"10.1007/s10489-025-06423-3","url":null,"abstract":"<div><p>Quantitative trading strategy optimization in the complex and dynamic financial markets presents good challenges due to market non-stationarity, bounded rationality of participants, and the lack of adaptability in existing algorithms. To address these challenges, we propose a novel adaptive quantitative trading strategy optimization framework that seamlessly integrates meta reinforcement learning, cognitive game theory, and automated strategy generation. Our framework achieves superior adaptability, robustness, and profitability, with annualized returns of 51.9%, 49.3%, 46.5%, and 53.7% and Sharpe ratios of 2.37, 2.21, 2.08, and 2.45 in the Chinese, US, European, and Japanese stock markets, respectively, outperforming traditional methods and state-of-the-art machine learning algorithms. The maximum drawdowns are limited to -10.2%, -11.4%, -12.1%, and -10.8%, and the Sortino ratios reach 3.54, 3.28, 3.07, and 3.68, demonstrating effective downside risk management. However, challenges remain in terms of computational complexity, the need for more extensive out-of-sample validation, the incorporation of advanced NLP techniques, and the extension to other markets and asset classes. These limitations call for further research efforts. Overall, this research makes notable contributions to quantitative trading, meta reinforcement learning, and cognitive game theory, opening up new avenues for the development of adaptive, robust, and high-performing trading strategies.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-04-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143871331","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

AI-driven 5G IoT e-nose for whiskey classification ai驱动的5G物联网电子鼻用于威士忌分类

IF 3.4 2区计算机科学

Applied Intelligence Pub Date : 2025-04-24 DOI: 10.1007/s10489-025-06425-1

Jaume Segura-Garcia, Rafael Fayos-Jordan, Mohammad Alselek, Sergi Maicas, Miguel Arevalillo-Herraez, Enrique A. Navarro-Camba, Jose M. Alcaraz-Calero

{"title":"AI-driven 5G IoT e-nose for whiskey classification","authors":"Jaume Segura-Garcia, Rafael Fayos-Jordan, Mohammad Alselek, Sergi Maicas, Miguel Arevalillo-Herraez, Enrique A. Navarro-Camba, Jose M. Alcaraz-Calero","doi":"10.1007/s10489-025-06425-1","DOIUrl":"10.1007/s10489-025-06425-1","url":null,"abstract":"<div><p>The main contribution is the design, implementation and validation of a complete AI-driven electronic nose architecture to perform the classification of whiskey and acetones. This classification is of paramount important in the distillery production line of whiskey in order to predict the quality of the final product. In this work, we investigate the application of an e-nose (based on arrays of single-walled carbon nanotubes) to the distinction of two different substances, such as whiskey and acetone (as a subproduct of the distillation process), and discrimination of three different types of the same substance, such as three types of whiskies. We investigated different strategies to classify the odor data and provided a suitable approach based on random forest with accuracy of 99% and with inference times under 1.8 seconds. In the case of clearly different substances, as subproducts of the whiskey distillation process, the procedure presented achieves a high accuracy in the classification process, with an accuracy around 96%.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10489-025-06425-1.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143865450","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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}

引用次数: 0

Enhancing robust node classification via information competition: An improved adversarial resilience method for graph attacks 通过信息竞争增强鲁棒节点分类：一种改进的图攻击对抗弹性方法

IF 3.4 2区计算机科学

Applied Intelligence Pub Date : 2025-04-24 DOI: 10.1007/s10489-025-06478-2

Yong Huang, Yao Yang, Qiao Han, Xinling Guo, Yiteng Zhai, Baoping Cheng

{"title":"Enhancing robust node classification via information competition: An improved adversarial resilience method for graph attacks","authors":"Yong Huang, Yao Yang, Qiao Han, Xinling Guo, Yiteng Zhai, Baoping Cheng","doi":"10.1007/s10489-025-06478-2","DOIUrl":"10.1007/s10489-025-06478-2","url":null,"abstract":"<div><p>Graph neural networks (GNNs) demonstrate their effectiveness in facilitating node classification and a range of graph-based tasks. However, recent studies have revealed that GNNs can be vulnerable to various adversarial attacks. Despite various defense strategies, ranging from attack-agnostic defenses to attack-oriented defenses that have been proposed to mitigate the impact of adversarial attacks on graph data, effectively learning attack-agnostic graph representation remains an open challenge. This paper introduces a novel information Competition-based framework for Graph Neural Networks (i.e., <i>iC</i>-GNN, e.g., <i>iC</i>-GCN, <i>iC</i>-GAT, etc.) to enhance the robustness of GNNs against various adversarial attacks in node classifications. Through the use of graph reconstruction and low-rank approximation, our approach learns diversified graph representations to collaboratively perform node classifications. Meanwhile, mutual information constraints are utilized on different graph representations to ensure diversity and competition in graph features. The experimental results indicate that within the proposed framework, <i>iC</i>-GCN outperforms other graph defense frameworks in countering a wide range of targeted and non-targeted adversarial attacks in both evasion and poisoning training scenarios. Additionally, this concept has been extended to encompass other widely utilized GNN models like <i>iC</i>-GAT and <i>iC</i>-SAGE. All <i>iC</i>-GNN models demonstrate effective defense capabilities, demonstrating comparable resilience to adversarial attacks. This underscores the superiority and scalable nature of the <i>iC</i>-GNN framework, providing opportunities for a variety of graph learning applications.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143865479","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multi-step prediction method of temperature and humidity based on TCN-FECAM-iTransformer 基于tcn - fecam - ittransformer的温湿度多步预测方法

IF 3.4 2区计算机科学

Applied Intelligence Pub Date : 2025-04-24 DOI: 10.1007/s10489-025-06572-5

Zongxu Xie, Zhiqing Tao, Xianhong Xie, Yuan Rao, Ke Li, Wei Li, Jun Zhu

{"title":"Multi-step prediction method of temperature and humidity based on TCN-FECAM-iTransformer","authors":"Zongxu Xie, Zhiqing Tao, Xianhong Xie, Yuan Rao, Ke Li, Wei Li, Jun Zhu","doi":"10.1007/s10489-025-06572-5","DOIUrl":"10.1007/s10489-025-06572-5","url":null,"abstract":"<div><p>Greenhouses are a critical component of modern agriculture, facilitating crop growth and development, and accurate predictions of temperature and humidity are essential for mitigating crop diseases and optimizing the growth environment. However, short- and medium-term forecasts of temperature and humidity are challenging because of the complexity of greenhouse microclimates. This paper presents a hybrid model that integrates a frequency-enhanced channel attention mechanism optimized with a temporal convolutional network (TCN-FECAM) and an iTransformer. The model employs a cross-attention mechanism incorporating the advantages of the two models, and a 48-sequence sliding window strategy is used to ensure accurate multistep predictions of temperature and humidity over spans of 3 h to 24 h. The experimental results demonstrate that the TCN-FECAM-iTransformer model outperforms other models across diverse time scales, including GRU, LSTM, Informer, Autoformer, Crossformer, FAM-LSTM, and TPA-LSTM. Specifically, in temperature prediction, the model achieves R<sup>2</sup> coefficients of 0.979, 0.973, 0.968, and 0.953 and RMSE values of 0.657, 0.806, 0.923, and 1.126, for 3 h, 6 h, 12 h, and 24 h intervals, respectively. In humidity prediction, the model obtains R<sup>2</sup> coefficients of 0.976, 0.961, 0.947, and 0.939 and RMSE values of 1.805, 2.567, 3.132, and 3.451 for 3 h, 6 h, 12 h, and 24 h intervals, respectively. The model therefore exhibits reliable performance in predicting temperature and humidity in greenhouse environments, offering robust support for monitoring and early warnings in crop growth environments.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"55 10","pages":""},"PeriodicalIF":3.4,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143865449","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0