检索结果-内蒙古大学图书馆

Constructing and Evaluating Predictors for data-Driven PM2.5 Forecasting Models

international JOURNAL OF ENVIRONMENTAL RESEARCH 2025年第3期19卷 1-15页

作者： He, Ran-Ran Chen, Yu-Qiao Tian, Lei Shan, Lei Sang, Xiao-Shuang Anhui Univ Finance & Econ Sch Management Sci & Engn Bengbu 233030 Anhui Peoples R China Bengbu Bur Meteorol Bengbu 233000 Anhui Peoples R China

data-driven methods for air quality forecasting are beneficial to early-warning of air pollution, especially in small and medium-sized cities. In this study, the Prophet model is used for extracting time-related features, such as trend, annual cycles, and weekly cycles, from the PM2.5 series, and the gradient boosting regression trees (GBM) and long short-term memory networks (LSTM) are used to build forecasting models utilizing predictors including time-related features. All predictors are grouped into time-related features, antecedent air pollutant concentration, and meteorological parameters, and permutation feature importance (PFI) is used for measuring the importance of these feature groups. A series of forecasting experiments on a middle city in China are carried out, and the results are analyzed for a better understanding of the mechanisms of the forecasting skill. The result shows that time-related features extracted by Prophet can enhance the forecasting skill significantly, and this advantage is more apparent when the forecasting horizon is large. PM2.5 time series has short memory and a rapid response to changes in meteorological conditions, and only meteorological conditions in the previous 2 days have useful signals for forecasting. The rapid response and short memory of PM2.5 make LSTM show no advantage over GBM in this case study. We show that feature engineering is important for forecasting PM2.5, and the time-related features and antecedent pollutant concentrations can be seen as proxy variables for some variables that are difficult to obtain, such as pollutant emission and missing meteorological conditions.

关键词： Air quality machine learning Feature engineering data-driven model PM2.5

来源：评论

学校读者我要写书评

暂无评论

Dynamic Soft Contrastive learning for Time Series Anomaly Detection

Dynamic Soft Contrastive Learning for Time Series Anomaly De...

引用

2025 IEEE international conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Song, Yifan Liu, Yu Shu, Shaolong College of Electronics and Information Engineering Tongji University Shanghai China

ISBN: (纸本)9798350368741

Multivariate time series anomaly detection has been extensively studied in diverse domains. Within the horizon of unsupervised methods, density estimation is regarded as a promising direction. However, the anomalies that distribute closely with normal data in data space impact the desired density estimation results, thereby significantly affecting the discrimination of anomalies. In this work, to tackle this challenge, we propose Dynamic Soft Contrastive learning (DiSCo). DiSCo generates a representation that better captures the underlying correlations within the data distribution, providing a more compact, informative, and suitable description for the subsequent density estimation task. Furthermore, DiSCo employs a dynamic and learnable distance metric to measure similarities between inputs. The learned task-specific distance metric contributes to minimizing the intra-class distance and maximizing the inter-class distance, thereby assisting the generation of better representation. The efficacy of DiSCo is proved through extensive experiments conducted on four common anomaly detection datasets. © 2025 IEEE.

关键词： Anomaly Detection Contrastive learning Normalizing Flow Time Series Unsupervised learning

来源：评论

学校读者我要写书评

暂无评论

Artificial Intelligence and machine learning Approaches in Eco-Design for Additive Manufacturing: A Literature Review

Artificial Intelligence and Machine Learning Approaches in E...

引用

international Joint conference on Mechanics, Design engineering and Advanced Manufacturing, JCM 2024

作者： Musiari, Francesco Marconi, Marco Villazzi, Nicola Murgese, Luca Mandolini, Marco Gallozzi, Simone Favi, Claudio Università degli Studi della Tuscia Largo dell’Università Viterbo01100 Italy Università degli Studi di Parma Parco Area delle Scienze 181/A Parma43124 Italy Università Politecnica delle Marche Via Brecce Bianche 12 Ancona60131 Italy

ISBN: (纸本)9783031728280

Additive manufacturing (AM) is currently considered one of the most promising technologies for fostering innovation in the industrial sector, specifically in product design. Considering the non-subtractive nature of AM processes, in first approximation, they all seem to be environmentally sustainable since they lead to a null or at least reduced quantity of scraps. However, the peculiarities of such technologies, such as the specific input materials (e.g., metal powders, polymer filaments) or the low productivity, make the situation more complex, not so homogeneous in all the cases, and strongly dependent on the process and product parameters. In this complex scenario, the present paper aims to investigate the current state of the art regarding adopting Artificial Intelligence (AI) and machine learning (ML) toward the eco-design for AM. Following the PRISMA methodology, a set of 19 scientific papers were analysed. Papers were grouped according to the benefits (improvement of quality and reduction of energy, waste and defects) and design opportunities (definition of shapes and selection of materials) that AI and ML potentially provide. The discussion of the results presents the most relevant research and development opportunities for scholars and companies. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Smart manufacturing

来源：评论

学校读者我要写书评

暂无评论

Scalable VLSI Architectures for Pattern Matching in Serial data Stream

Scalable VLSI Architectures for Pattern Matching in Serial D...

引用

2025 international conference on Electronics, Information, and Communication, ICEIC 2025

作者： Kong, Byeong Yong Institute of IT Convergence Technology Kongju National University Division of Electrical Electronic and Control Engineering Cheonan Korea Republic of

ISBN: (纸本)9798331510756

In this paper, a pair of regularized and scalable very large scale integration (VLSI) architectures are presented for matching patterns in a serial data stream. One is grounded on a shift register (SR), whereas the other is based on a finite state machine (FSM), both of which are very easy to constitute for arbitrary lengths of pattern. While the architectures for parallel pattern matching have been extensively studied in the literature, those for serially incoming data have been rare. To suggest a guideline for efficient implementation under such circumstances, the pros and cons of the proposed architectures are analyzed in various aspects. © 2025 IEEE.

关键词： Network intrusion

来源：评论

学校读者我要写书评

暂无评论

Magneto Inductive (MI) Channel Variables Prediction Through machine learning Linear Regression Method, for Underwater and Underground WSNs

引用

IEEE ACCESS 2025年 13卷 33124-33137页

作者： Muhammad, Aman Li, Fuzhong Mohsan, Syed Agha Hassnain Khan, Zahid Ullah Khan, Waqar Khan, Sajid Ullah Han, Ziqi Khan, Faheem Shanxi Agr Univ Coll Agr Engn Jinzhong 030801 Peoples R China Shanxi Agr Univ Sch Software Jinzhong 030801 Peoples R China Zhejiang Univ Ocean Coll Zhoushan 310027 Zhejiang Peoples R China Harbin Engn Univ Coll Informat & Commun Engn Harbin 150001 Peoples R China Fuzhou Univ Int Studies & Trade Sch Big Data Fuzhou 350200 Fujian Peoples R China Prince Sattam bin Abdulaziz Univ Coll Comp Engn & Sci Dept Informat Syst Al Kharj 16278 Saudi Arabia Univ Engn & Technol Peshawar Dept Elect Engn Peshawar 610636 Pakistan

In today's digital era, where information flows seamlessly and is readily available and accessible. However, these information and communication systems are highly dependent on the robustness and adaptability of communication links, which must endure the dynamic and often unpredictable changes occurring in the communication environment, especially in underwater and underground communication. As technology advances and demands for reliable data transfer grows, the resilience of these links becomes increasingly critical to ensure an uninterrupted access to information. For this we will be analyzing a new paradigm in Wireless Sensor Networks (WSNs), i.e. Magneto Inductive (MI) communication. Which has the capability to provide robustness to the communication link as well as the security of the information can also be controlled, by utilizing the directionality factor of the created magnetic flux around the MI tri-directional (MI-TD) coils. In this study we will examine MI communication variable factors, which have effects on the output of the MI channel's response, such as frequency of applied excitation current, number of turns, radius of the communicating coils, and changing communication distance between nodes. Than we will use a linear regression machine learning (ML) model to predict the values of the aforementioned variables, which will help to predict the MI channel's response in the form of Path loss/Bit Error Rate (BER). The model that we have created for the prediction purpose, archives an estimation accuracy of 92%.

关键词： Coils Wireless sensor networks Magnetic flux Magnetic fields Wireless communication Magnetic field measurement Receivers Predictive models Media Magnetic domains Magneto inductive communication machine learning linear regression information security magnetic flux magneto inductive channel wireless sensor networks

来源：评论

学校读者我要写书评

暂无评论

Convolutional-and Deep learning-Based Techniques for Time Series Ordinal Classification

引用

IEEE TRANSACTIONS ON CYBERNETICS 2025年第2期55卷 537-549页

作者： Ayllon-Gavilan, Rafael Guijo-Rubio, David Gutierrez, Pedro Antonio Bagnall, Anthony Hervas-Martinez, Cesar Inst Maimonides Invest Biomed Cordoba Dept Clin Epidemiol Res Primary Care Cordoba 14004 Spain Univ Cordoba Dept Comp Sci & Numer Anal Cordoba 14071 Spain Univ Southampton Dept Elect & Comp Sci Southampton SO17 1BJ England

Time-series classification (TSC) covers the supervised learning problem where input data is provided in the form of series of values observed through repeated measurements over time, and whose objective is to predict the category to which they belong. When the class values are ordinal, classifiers that take this into account can perform better than nominal classifiers. Time-series ordinal classification (TSOC) is the field bridging this gap, yet unexplored in the literature. There are a wide range of time-series problems showing an ordered label structure, and TSC techniques that ignore the order relationship discard useful information. Hence, this article presents the first benchmarking of TSOC methodologies, exploiting the ordering of the target labels to boost the performance of current TSC state of the art. Both convolutional-and deep-learning-based methodologies (among the best performing alternatives for nominal TSC) are adapted for TSOC. For the experiments, a selection of $29$ ordinal problems has been made. In this way, this article contributes to the establishment of the state of the art in TSOC. The results obtained by ordinal versions are found to be significantly better than current nominal TSC techniques in terms of ordinal performance metrics, outlining the importance of considering the ordering of the labels when dealing with this kind of problems.

关键词： Time series analysis Kernel Feature extraction Accuracy Transformers Trajectory Training Taxonomy Cybernetics Convolution Ordinal classification time-series analysis time-series classification (TSC) time-series machine learning (ML)

来源：评论

学校读者我要写书评

暂无评论

Analyzing COVID-19 sentiments on Twitter: A machine learning approach 4th

Analyzing COVID-19 sentiments on Twitter: A machine learning...

引用

4th international conference on Computational Methods in Science and Technology, ICCMST 2024

作者： Aishwarya, V. Sanmuga Priyan, K. Vyshnavi, R. Azhagiri, M. Sathya Priya, S. SRM Institute of Science and Technology Chennai India

ISBN: (纸本)9781032911571

Throughout the outbreak, COVID-19 has been a hot issue on Twitter, providing a forum for people to voice their ideas and perspectives. Unfortunately, there isn’t a thorough rundown of sentiment analysis in Twitter discussions around COVID-19. By analyzing sentiment analysis utilizing COVID-19 Twitter data from both a machine learning and a behavioral and social science standpoint, this article fills this gap. Forty papers examining public attitudes were found in a study of the literature conducted between October 2019 and January 2022. Higher accuracy is shown by ensemble models, especially those that use BERT and RoBERTa. These models do not provide exact outcomes, despite their theoretical promise. Based on the COVIDSENTI dataset, the research emphasizes the need for government action on social media to counter false information. It highlights that, despite the continual examination of datasets, proactive public health actions are crucial in lowering unfavorable sentiments during a pandemic. © 2025 the Author(s).

关键词： Tweets

来源：评论

学校读者我要写书评

暂无评论

FedFLD: Heterogeneous Federated learning via Forget-Less Distillation

FedFLD: Heterogeneous Federated Learning via Forget-Less Dis...

引用

2025 IEEE international conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Yi, Xiaoyang Zhang, Jian Chen, Jing Bao, Yuru Xing, Lingkai College of Cyber Science Nankai University China College of Computer Science Nankai University China Tianjin Key Laboratory of Network and Data Security Technology Tianjin China

ISBN: (纸本)9798350368741

Federated learning, as a distributed machine learning paradigm, enhances privacy protection but faces the challenge of heterogeneity. data-free knowledge distillation (DFKD) methods attempt to overcome this challenge by using a generator to synthesize samples for fine-tuning a global model. However, these methods often suffer from significant shifts in output distribution, leading to catastrophic forgetting. To tackle these issues, we propose FedFLD, a novel federated forget-less distillation framework that mitigates catastrophic forgetting in DFKD while addressing the heterogeneity challenge. Specifically, FedFLD guides the generator's training from three key aspects and constrains the output distribution with an elastic weight consolidation penalty term. By synthesizing diverse samples from different perspectives through additional generator updates, FedFLD facilitates effective knowledge distillation from local models to the global model. Additionally, the global model is further optimized via the heterogeneity fine-tuning process, mitigating the bias from heterogeneity and resulting in a more expressive and robust model. © 2025 IEEE.

关键词： Catastrophic Forgetting data-free Knowledge Distillation Federated learning Heterogeneity

来源：评论

学校读者我要写书评

暂无评论

Comparison of Regressors Applied to 3D Motion Marker data of a Historic Building Prototype Subjected to Shake Table Tests 5th

Comparison of Regressors Applied to 3D Motion Marker Data of...

引用

5th international conference on Protection of Historical Constructions, PROHITECH 2025

作者： Palumbo, Domenico Ormando, Chiara Roselli, Ivan ENEA Casaccia Research Centre Rome Italy

ISBN: (纸本)9783031873157

In the present study the performance of several machine learning regression techniques for the analysis of displacement data acquired in dynamic identification shaking table tests to assess a building’s prototype state of damage were compared. The displacement data of a typical Italian historic masonry building prototype were recorded by a passive 3D motion capture system. These recorded displacement data were analyzed by calculating a widely accepted Damage Index (DI) based on the decay of the first modal frequency of the tested structure. To such purpose, a conventional Frequency Response Function (FRF) algorithm for modal analysis was adopted. Then, several regressors, commonly used in machine learning applications, were applied to cepstral coefficients extracted from time-series segmentation of the 3D motion markers signals to predict the DI values. To assess the quality of the tested regressors, Coefficient of determination (R2) and Relative Percent Deviation (RPD) parameters were evaluated. Different lengths of time-series segments were also explored and the results were statistically compared to highlight the differences between regressors. Excellent results were obtained in terms of R2 (0.98) and RPD (6.87) with a weak dependence on the time partitioning performed on the data with optimal results for 4 or more seconds. Moreover, Stacked Regressor, Decision Tree Bagging Regressor and XGB Regressor were found as the best overall regressors. The results obtained demonstrated that machine learning methods can be effectively utilized to predict the level of damage of historic masonry buildings with very high accuracy. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Masonry materials

来源：评论

学校读者我要写书评

暂无评论

Design of optimization model for tourism programme based on user profiling and factorization machine 2

Design of optimization model for tourism programme based on ...

引用

2nd international conference on Big data, Computational Intelligence, and Applications, BDCIA 2024

作者： Luo, Zhihui School of Foreign Language and International Business Guangdong Mechanical and Electrical Polytechnic Guangzhou510515 China

ISBN: (纸本)9781510689053

With the advent of big data analytics and 5G technology, the tourism industry has seen significant advancements, particularly in the development of personalized recommendation systems. In order to solve the problems of information overload as well as user personalised recommendation, the study proposes a tourism solution optimisation model based on user portrait and factor decomposition machine. Firstly, a user portrait containing static and dynamic attributes is constructed through data collection and preprocessing techniques, and the attraction features are clustered by combining with a clustering algorithm. The predictive ability of the model is enhanced by adding an interaction term of similarity between travel interests and attraction features. According to the experimental findings, the optimized model performs better on both the training and test sets than the classic model in terms of hit rate and normalized discount cumulative gain. When the number of recommendations is set to 5, the model achieves a hit rate and normalised discount cumulative gain of 0.5216 and 0.3389 on the training dataset, and the test dataset results are 0.5198 and 0.3309, *** recommendations is increased to 10. The hit rate and normalised discount cumulative gain of the proposed model of the study on the training and test datasets are still significantly higher than the other two compared models. The model proposed by the research institute not only improves the accuracy of recommendations, but also has broader application potential in personalized digital services. © 2025 SPIE.

关键词： 5G mobile communication systems

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：