Applied Soft Computing最新文献_第9页

A least-squares framework for developing interval type-2 fuzzy semantics 开发区间 2 型模糊语义的最小二乘法框架

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2024-10-05 DOI: 10.1016/j.asoc.2024.112293

The developing of IT2 fuzzy semantics is critical for computing with words (CWW), but exiting approaches lack flexibility and fail to adapt user’s diversified demands. The study proposes a least-squares framework for designing CWW encoders that construct interval type-2 fuzzy sets (IT2 FSs) to represent the semantic meanings of linguistic words. In the least-squares framework, an CWW encoder is characterized by two elements: an intra-uncertain semantic mapping and an inter-uncertain semantic family. The intra-uncertain semantic mapping transforms data intervals into type-1 fuzzy sets (T1 FSs), then an optimal IT2 FS is derived from the inter-uncertain semantic family using the least-squares method. Furthermore, several intra-uncertain semantic mappings are introduced, and a compatibility measure is defined to facilitate model selection. The least-squares framework benefits from the flexible selection of intra-uncertain semantic mappings and least-squares optimization-based construction of IT2 FSs. In experiments, the least-squares framework is applied to handle real-world online survey data and the large-scale online review data set of a Chinese life service review site, Dianping.com. Compared to the enhanced interval approach and the Hao-Mendel approach, the least-squares framework shows its favorable efficiency in experiments and statistical tests, and adapts to user-defined intra- and inter-uncertain semantic families.

开发 IT2 模糊语义对于用词计算（CWW）至关重要，但现有方法缺乏灵活性，无法适应用户的多样化需求。本研究提出了一个最小二乘法框架，用于设计构建区间 2 型模糊集（IT2 FS）来表示语词语义的 CWW 编码器。在最小二乘法框架中，CWW 编码器由两个要素构成：内部不确定语义映射和内部不确定语义族。内部不确定语义映射将数据区间转换为类型-1 模糊集（T1 FS），然后使用最小二乘法从内部不确定语义族推导出最优的 IT2 FS。此外，还引入了几种内部不确定语义映射，并定义了一种相容性度量，以方便模型选择。最小二乘法框架得益于不确定语义内映射的灵活选择和基于最小二乘法优化的 IT2 FSs 构建。在实验中，最小二乘框架被用于处理真实世界的在线调查数据和中国生活服务点评网站大众点评网的大规模在线点评数据集。与增强区间法和郝-门德尔法相比，最小二乘法框架在实验和统计检验中表现出良好的效率，并能适应用户定义的内部和内部不确定语义族。

{"title":"A least-squares framework for developing interval type-2 fuzzy semantics","authors":"","doi":"10.1016/j.asoc.2024.112293","DOIUrl":"10.1016/j.asoc.2024.112293","url":null,"abstract":"<div><div>The developing of IT2 fuzzy semantics is critical for computing with words (CWW), but exiting approaches lack flexibility and fail to adapt user’s diversified demands. The study proposes a least-squares framework for designing CWW encoders that construct interval type-2 fuzzy sets (IT2 FSs) to represent the semantic meanings of linguistic words. In the least-squares framework, an CWW encoder is characterized by two elements: an intra-uncertain semantic mapping and an inter-uncertain semantic family. The intra-uncertain semantic mapping transforms data intervals into type-1 fuzzy sets (T1 FSs), then an optimal IT2 FS is derived from the inter-uncertain semantic family using the least-squares method. Furthermore, several intra-uncertain semantic mappings are introduced, and a compatibility measure is defined to facilitate model selection. The least-squares framework benefits from the flexible selection of intra-uncertain semantic mappings and least-squares optimization-based construction of IT2 FSs. In experiments, the least-squares framework is applied to handle real-world online survey data and the large-scale online review data set of a Chinese life service review site, <em>Dianping.com</em>. Compared to the enhanced interval approach and the Hao-Mendel approach, the least-squares framework shows its favorable efficiency in experiments and statistical tests, and adapts to user-defined intra- and inter-uncertain semantic families.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":null,"pages":null},"PeriodicalIF":7.2,"publicationDate":"2024-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142441286","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}

引用次数: 0

Efficient prediction uncertainty quantification in dam behavior monitoring with attention-based sequence-to-sequence learning 利用基于注意力的序列到序列学习在大坝行为监测中高效量化预测不确定性

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2024-10-05 DOI: 10.1016/j.asoc.2024.112321

Displacement is an intuitive monitoring indicator of dam structural behavior. Conventional deterministic modeling methods frequently disregard the inherent uncertainty associated with monitoring data. This study aims to quantify such uncertainties within dam displacement predictions by augmenting a sequence-to-sequence structure with the novel recurrent unit, termed as tiny gated unit (TGU), attention mechanism, as well as quantile regression, resulting in the Att-S2STQ model. Specifically, TGU adopts only one gate to identify nonlinearity in data sequences, while Bahdanau attention mechanism dynamically assigns weights to different parts of the input sequence. The inclusion of quantile regression allows Att-S2STQ to produce prediction intervals (PIs), from which the probability density functions (PDFs) and cumulative distribution functions (CDFs) are further derived via kernel density estimation. PIs, PDFs, and CDFs jointly serve to reveal the prediction uncertainty. The effectiveness of the model is illustrated through comparative experiments using real-world dam monitoring datasets. Results indicate that the proposed model outperforms traditional methods in point, interval, and probability predictions, while also having a simpler structure and faster training. The superiority of both accuracy and efficiency makes it a valuable tool for dam management, aiding in data-driven decision-making and enhancing operational safety.

位移是大坝结构行为的直观监测指标。传统的确定性建模方法经常忽略与监测数据相关的内在不确定性。本研究旨在通过新颖的递归单元（称为微小门控单元（TGU））、注意机制以及量子回归来增强序列到序列结构，从而量化大坝位移预测中的不确定性，最终形成 Att-S2STQ 模型。具体来说，TGU 只采用一个门来识别数据序列中的非线性，而 Bahdanau 注意机制则动态地为输入序列的不同部分分配权重。加入量子回归后，Att-S2STQ 可以生成预测区间（PI），并通过核密度估计进一步推导出概率密度函数（PDF）和累积分布函数（CDF）。PI、PDF 和 CDF 共同揭示了预测的不确定性。通过使用真实世界的大坝监测数据集进行对比实验，说明了该模型的有效性。结果表明，所提出的模型在点预测、区间预测和概率预测方面都优于传统方法，而且结构更简单，训练更快速。准确性和效率的双重优势使其成为大坝管理的重要工具，有助于数据驱动决策和提高运行安全。

{"title":"Efficient prediction uncertainty quantification in dam behavior monitoring with attention-based sequence-to-sequence learning","authors":"","doi":"10.1016/j.asoc.2024.112321","DOIUrl":"10.1016/j.asoc.2024.112321","url":null,"abstract":"<div><div>Displacement is an intuitive monitoring indicator of dam structural behavior. Conventional deterministic modeling methods frequently disregard the inherent uncertainty associated with monitoring data. This study aims to quantify such uncertainties within dam displacement predictions by augmenting a sequence-to-sequence structure with the novel recurrent unit, termed as tiny gated unit (TGU), attention mechanism, as well as quantile regression, resulting in the Att-S2STQ model. Specifically, TGU adopts only one gate to identify nonlinearity in data sequences, while Bahdanau attention mechanism dynamically assigns weights to different parts of the input sequence. The inclusion of quantile regression allows Att-S2STQ to produce prediction intervals (PIs), from which the probability density functions (PDFs) and cumulative distribution functions (CDFs) are further derived via kernel density estimation. PIs, PDFs, and CDFs jointly serve to reveal the prediction uncertainty. The effectiveness of the model is illustrated through comparative experiments using real-world dam monitoring datasets. Results indicate that the proposed model outperforms traditional methods in point, interval, and probability predictions, while also having a simpler structure and faster training. The superiority of both accuracy and efficiency makes it a valuable tool for dam management, aiding in data-driven decision-making and enhancing operational safety.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":null,"pages":null},"PeriodicalIF":7.2,"publicationDate":"2024-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142427735","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}

引用次数: 0

A two-phase adaptive large neighborhood search algorithm for the electric location routing problem with range anxiety 针对具有范围焦虑的电力位置路由问题的两阶段自适应大邻域搜索算法

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2024-10-05 DOI: 10.1016/j.asoc.2024.112323

In response to environmental pollution and global warming, electric vehicles (EVs) have been widely applied in supply chain, and lots of researches have been focused on the electric location routing problem (ELRP) to optimize the location of battery-swapping stations (BSSs). However, few studies have simultaneously considered the EV’s collaborative behavior and range anxiety in ELRP. Therefore, this study focuses on formulating an ELRP with battery-swapping stations considering both collaborative behavior and range anxiety. The model takes into account the collaboration between the supply and demand sides to share supply chain resources effectively. The detour probability function is employed to handle the uncertainty caused by drivers’ range anxiety during the trip, which can be evaluated through a proposed range anxiety function. In addition, a new two-dimensional matrix-based solution representation is proposed to explore ELRP optimal solutions intuitively and efficiently. To solve the proposed model effectively, a two-phase adaptive large neighborhood search (TALNS) algorithm that integrates the adaptive large neighborhood search (ALNS) algorithm and the extended binary particle swarm optimization (EBPSO) algorithm is proposed. In the EBPSO algorithm, a new position update mechanism and local search strategy are used to strengthen the local search ability. In the ALNS algorithm, multiple destroy and repair operators are developed for the proposed model. Further, to validate the effectiveness and performance of the proposed algorithm, comparison experiments with the Gurobi optimizer and other baseline heuristic algorithms are conducted.

为了应对环境污染和全球变暖，电动汽车（EV）已被广泛应用于供应链中，许多研究都集中在电动汽车位置路由问题（ELRP）上，以优化电池更换站（BSS）的位置。然而，很少有研究在 ELRP 中同时考虑电动汽车的协作行为和续航焦虑。因此，本研究侧重于制定一个同时考虑协作行为和续航焦虑的有电池更换站的 ELRP。该模型考虑了供需双方的协作，以有效共享供应链资源。采用绕行概率函数来处理驾驶员在行驶过程中因里程焦虑而产生的不确定性，并通过提出的里程焦虑函数对其进行评估。此外，还提出了一种新的基于二维矩阵的求解表示法，以直观高效地探索 ELRP 的最优解。为有效求解所提出的模型，提出了一种两阶段自适应大邻域搜索（TALNS）算法，该算法集成了自适应大邻域搜索（ALNS）算法和扩展二元粒子群优化（EBPSO）算法。在 EBPSO 算法中，采用了新的位置更新机制和局部搜索策略，以加强局部搜索能力。在 ALNS 算法中，为所提出的模型开发了多个销毁和修复算子。此外，为了验证所提算法的有效性和性能，还与 Gurobi 优化器和其他基线启发式算法进行了对比实验。

{"title":"A two-phase adaptive large neighborhood search algorithm for the electric location routing problem with range anxiety","authors":"","doi":"10.1016/j.asoc.2024.112323","DOIUrl":"10.1016/j.asoc.2024.112323","url":null,"abstract":"<div><div>In response to environmental pollution and global warming, electric vehicles (EVs) have been widely applied in supply chain, and lots of researches have been focused on the electric location routing problem (ELRP) to optimize the location of battery-swapping stations (BSSs). However, few studies have simultaneously considered the EV’s collaborative behavior and range anxiety in ELRP. Therefore, this study focuses on formulating an ELRP with battery-swapping stations considering both collaborative behavior and range anxiety. The model takes into account the collaboration between the supply and demand sides to share supply chain resources effectively. The detour probability function is employed to handle the uncertainty caused by drivers’ range anxiety during the trip, which can be evaluated through a proposed range anxiety function. In addition, a new two-dimensional matrix-based solution representation is proposed to explore ELRP optimal solutions intuitively and efficiently. To solve the proposed model effectively, a two-phase adaptive large neighborhood search (TALNS) algorithm that integrates the adaptive large neighborhood search (ALNS) algorithm and the extended binary particle swarm optimization (EBPSO) algorithm is proposed. In the EBPSO algorithm, a new position update mechanism and local search strategy are used to strengthen the local search ability. In the ALNS algorithm, multiple destroy and repair operators are developed for the proposed model. Further, to validate the effectiveness and performance of the proposed algorithm, comparison experiments with the Gurobi optimizer and other baseline heuristic algorithms are conducted.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":null,"pages":null},"PeriodicalIF":7.2,"publicationDate":"2024-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142437897","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}

引用次数: 0

A cognitive load assessment method for fighter cockpit human-machine interface based on integrated multi-criteria decision making 基于综合多标准决策的战斗机驾驶舱人机界面认知负荷评估方法

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2024-10-05 DOI: 10.1016/j.asoc.2024.112287

The primary interface for communication between pilots and aircraft systems is the fighter cockpit Human-Machine Interface (HMI). Since reduced cognitive load ensures that pilots will operate the aircraft safely and effectively, it is critical to evaluate pilot cognitive load during HMI. A cognitive load assessment method, based on multi-criteria decision-making, is proposed to accurately quantify the relationship between the HMI and the cognitive load in the fighter’s cockpit. Firstly, based on the integrated Multi-Criteria Decision-Making (MCDM) method, the Step-wise Weight Assessment Ratio Analysis (SWARA) and MEthod based on the Removal Effects of Criteria (MEREC) methods are used to assign subjective and objective weights, respectively. Moreover, the Combined Compromise Solution (CoCoSo) method is applied to rank the scenarios to establish a cognitive load assessment model for the cockpit HMI of the fighter jet. Secondly, an evaluation standard system of fighter cockpit HMI is proposed, drawn upon multiple sets of eye-movement criteria and subjective assessment criteria. Moreover, cognitive load experiments of fighter cockpit HMI are conducted using eye-tracking technology to get the objective physiological cognitive data as well as the subjective assessment data of the subjects. Consequently, the parameter data sets of the eye-movement criteria and the subjective criteria for the evaluation of cognitive load are obtained. The proposed method is applied to analyze the cognitive load assessment of a fighter jet cockpit HMI layout. This application aims to verify the effectiveness of the assessment method in evaluating the cognitive load of the HMI layout. Through sensitivity and comparison analyses, the model was further verified to have excellent robustness and applicability for cognitive load assessment. The advantages of this method can be seen through the comparison, which is that it has higher discriminability when assessing the degree of cognitive load. At the same time, it has higher flexibility in dealing with complex and ambiguous cognitive load assessment information.

战斗机驾驶舱人机界面（HMI）是飞行员与飞机系统之间进行交流的主要界面。由于减少认知负荷可确保飞行员安全有效地操作飞机，因此评估人机界面期间飞行员的认知负荷至关重要。本文提出了一种基于多标准决策的认知负荷评估方法，以准确量化人机界面与战斗机驾驶舱认知负荷之间的关系。首先，在综合多标准决策（MCDM）方法的基础上，采用分步权重评估比率分析法（SWARA）和基于标准移除效应的方法（MEREC）分别分配主观权重和客观权重。此外，还采用综合折衷方案（CoCoSo）法对场景进行排序，以建立战斗机驾驶舱人机界面的认知负荷评估模型。其次，根据多套眼动标准和主观评价标准，提出了战斗机驾驶舱人机界面的评价标准体系。此外，还利用眼动跟踪技术对战斗机驾驶舱人机界面进行了认知负荷实验，以获得客观的生理认知数据和受试者的主观评价数据。因此，获得了眼动标准和认知负荷评估主观标准的参数数据集。所提出的方法被用于分析战斗机驾驶舱人机界面布局的认知负荷评估。该应用旨在验证评估方法在评估人机界面布局认知负荷方面的有效性。通过灵敏度和对比分析，进一步验证了该模型在认知负荷评估方面具有出色的稳健性和适用性。通过对比可以看出该方法的优势，即在评估认知负荷程度时具有更高的可辨别性。同时，它在处理复杂和模糊的认知负荷评估信息时具有更高的灵活性。

{"title":"A cognitive load assessment method for fighter cockpit human-machine interface based on integrated multi-criteria decision making","authors":"","doi":"10.1016/j.asoc.2024.112287","DOIUrl":"10.1016/j.asoc.2024.112287","url":null,"abstract":"<div><div>The primary interface for communication between pilots and aircraft systems is the fighter cockpit Human-Machine Interface (HMI). Since reduced cognitive load ensures that pilots will operate the aircraft safely and effectively, it is critical to evaluate pilot cognitive load during HMI. A cognitive load assessment method, based on multi-criteria decision-making, is proposed to accurately quantify the relationship between the HMI and the cognitive load in the fighter’s cockpit. Firstly, based on the integrated Multi-Criteria Decision-Making (MCDM) method, the Step-wise Weight Assessment Ratio Analysis (SWARA) and MEthod based on the Removal Effects of Criteria (MEREC) methods are used to assign subjective and objective weights, respectively. Moreover, the Combined Compromise Solution (CoCoSo) method is applied to rank the scenarios to establish a cognitive load assessment model for the cockpit HMI of the fighter jet. Secondly, an evaluation standard system of fighter cockpit HMI is proposed, drawn upon multiple sets of eye-movement criteria and subjective assessment criteria. Moreover, cognitive load experiments of fighter cockpit HMI are conducted using eye-tracking technology to get the objective physiological cognitive data as well as the subjective assessment data of the subjects. Consequently, the parameter data sets of the eye-movement criteria and the subjective criteria for the evaluation of cognitive load are obtained. The proposed method is applied to analyze the cognitive load assessment of a fighter jet cockpit HMI layout. This application aims to verify the effectiveness of the assessment method in evaluating the cognitive load of the HMI layout. Through sensitivity and comparison analyses, the model was further verified to have excellent robustness and applicability for cognitive load assessment. The advantages of this method can be seen through the comparison, which is that it has higher discriminability when assessing the degree of cognitive load. At the same time, it has higher flexibility in dealing with complex and ambiguous cognitive load assessment information.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":null,"pages":null},"PeriodicalIF":7.2,"publicationDate":"2024-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142528085","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}

引用次数: 0

Preference learning based on adaptive graph neural network for multi-criteria decision support 基于自适应图神经网络的偏好学习，用于多标准决策支持

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2024-10-05 DOI: 10.1016/j.asoc.2024.112312

Intelligent decision-making assists decision-makers (DMs) in making choices through data analysis, model prediction, and automated processes. Central to this field are two key concepts: multi-criteria decision making (MCDM) and preference learning (PL). While MCDM and PL both aim to develop decision models that rank alternatives based on observed or revealed preferences, they diverge in focus. MCDM concentrates on the DMs' perspectives, whereas PL emphasizes model-driven approaches. This divergence presents significant challenges in integrating these methodologies, particularly in ensuring the integrated method remains scalable and interpretable amidst the complexity of decision scenarios. To bridge this gap, our study introduces the use of graph structures to frame decision problems and proposes a novel PL method employing graph neural network (GNN) for multi-criteria decision support. This method is anchored in the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) technique and combines an adaptive GNN model with a weight determination model. The GNN model updates embeddings from the alternative's criteria and category features, utilizing an attention mechanism to adaptively assess their importance. Concurrently, the weight determination model contains a weight neural network module to set objective criterion weights and a game theory-based module for calculating combined criterion weights. The method not only inherits the interpretability and intuitive appeal of decision models but also leverages the computational efficiency and high accuracy of machine learning. In experiments conducted on benchmark datasets, our method exhibits significant performance improvements, especially in ranking-related evaluation metrics, outperforming the best baseline by 5.78 %.

智能决策通过数据分析、模型预测和自动化流程帮助决策者（DMs）做出选择。这一领域的核心是两个关键概念：多标准决策（MCDM）和偏好学习（PL）。虽然多标准决策（MCDM）和偏好学习（PL）都旨在开发决策模型，根据观察到的或揭示的偏好对备选方案进行排序，但它们的侧重点有所不同。MCDM 侧重于管理者的视角，而 PL 则强调模型驱动的方法。这种分歧给这些方法的整合带来了巨大挑战，尤其是如何确保整合后的方法在复杂的决策场景中保持可扩展性和可解释性。为了弥补这一差距，我们的研究引入了图结构来构建决策问题，并提出了一种采用图神经网络（GNN）的新型 PL 方法，用于多标准决策支持。该方法以 "通过与理想解决方案相似度排序偏好技术"（TOPSIS）为基础，将自适应 GNN 模型与权重确定模型相结合。GNN 模型从备选方案的标准和类别特征中更新嵌入，利用注意力机制自适应地评估其重要性。同时，权重确定模型包含一个权重神经网络模块，用于设置客观标准权重，以及一个基于博弈论的模块，用于计算综合标准权重。该方法不仅继承了决策模型的可解释性和直观性，还利用了机器学习的计算效率和高准确性。在基准数据集上进行的实验中，我们的方法表现出显著的性能改进，尤其是在与排名相关的评价指标方面，比最佳基线高出 5.78%。

{"title":"Preference learning based on adaptive graph neural network for multi-criteria decision support","authors":"","doi":"10.1016/j.asoc.2024.112312","DOIUrl":"10.1016/j.asoc.2024.112312","url":null,"abstract":"<div><div>Intelligent decision-making assists decision-makers (DMs) in making choices through data analysis, model prediction, and automated processes. Central to this field are two key concepts: multi-criteria decision making (MCDM) and preference learning (PL). While MCDM and PL both aim to develop decision models that rank alternatives based on observed or revealed preferences, they diverge in focus. MCDM concentrates on the DMs' perspectives, whereas PL emphasizes model-driven approaches. This divergence presents significant challenges in integrating these methodologies, particularly in ensuring the integrated method remains scalable and interpretable amidst the complexity of decision scenarios. To bridge this gap, our study introduces the use of graph structures to frame decision problems and proposes a novel PL method employing graph neural network (GNN) for multi-criteria decision support. This method is anchored in the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) technique and combines an adaptive GNN model with a weight determination model. The GNN model updates embeddings from the alternative's criteria and category features, utilizing an attention mechanism to adaptively assess their importance. Concurrently, the weight determination model contains a weight neural network module to set objective criterion weights and a game theory-based module for calculating combined criterion weights. The method not only inherits the interpretability and intuitive appeal of decision models but also leverages the computational efficiency and high accuracy of machine learning. In experiments conducted on benchmark datasets, our method exhibits significant performance improvements, especially in ranking-related evaluation metrics, outperforming the best baseline by 5.78 %.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":null,"pages":null},"PeriodicalIF":7.2,"publicationDate":"2024-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142427827","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}

引用次数: 0

Automated EEG-based language detection using directed quantum pattern technique 利用定向量子模式技术进行基于脑电图的语言自动检测

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2024-10-05 DOI: 10.1016/j.asoc.2024.112301

Electroencephalogram (EEG) signals contain complex useful information about brain activities. These EEG signals are noisy, highly varying and nonstationary in nature. Hence, extracting meaningful information from these signals is challenging. The existing machine learning systems struggle to capture the minute changes from the signals and yield high performance.

This study introduces a novel quantum-inspired feature extraction technique called Directed Quantum Pattern (DQP), designed to address these challenges by using a lattice structure to capture directional binary features. These directions (paths) are computed using a maximum function providing a dynamic and adaptive feature representation.

This paper presents a novel DQP-LangNet developed using DQP for automated classification of two- languages using EEG signals. We have proposed a hybrid approach, combining DQP, statistical features, and multi-level discrete wavelet transform (MDWT) to extract salient features similar to the deep learning approach. The EEG dataset consisting of 14 channels, produces 7 feature vectors per channel, yielding 98 feature vectors. Neighborhood component analysis and Chi-square (Chi2) feature selection approaches generated 196 feature vectors.

In addition to the innovative feature extraction a new classification structure called “t” is proposed k-nearest neighbor (tkNN) and support vector machine (tSVM) classifiers are employed. Using the proposed tkNN and tSVM classifiers, 392 (=196×2) classifier-based outcomes are obtained. To further improve classification performance, we applied the iterative majority voting (IMV) technique to automatically select the best result.

Our DQP-based model achieved a classification accuracy of 95.68 %using EEG language dataset with leave-one-subject-out (LOSO) cross-validation strategy. Also, an explainable feature engineering (XFE) structure of DQP-LangNet is employed to obtain channel-specific explainable results. Our proposed DQP-LangNet model can be employed for other applications in neuroscience.

脑电图（EEG）信号包含有关大脑活动的复杂有用信息。这些脑电信号具有噪声大、变化大和非稳态的特点。因此，从这些信号中提取有意义的信息具有挑战性。本研究介绍了一种名为定向量子模式（DQP）的新型量子启发特征提取技术，旨在通过使用晶格结构捕捉方向性二进制特征来应对这些挑战。这些方向（路径）使用最大值函数计算，提供了一种动态和自适应的特征表示。本文介绍了一种使用 DQP 开发的新型 DQP-LangNet，用于使用脑电信号对两种语言进行自动分类。我们提出了一种混合方法，将 DQP、统计特征和多级离散小波变换 (MDWT) 结合起来，以提取与深度学习方法类似的突出特征。脑电图数据集由 14 个通道组成，每个通道产生 7 个特征向量，共产生 98 个特征向量。除了创新的特征提取外，还提出了一种名为 "t "的新分类结构，即 k-近邻（tkNN）和支持向量机（tSVM）分类器。利用提出的 tkNN 和 tSVM 分类器，得到了 392 (=196×2) 个基于分类器的结果。为了进一步提高分类性能，我们采用了迭代多数投票（IMV）技术来自动选择最佳结果。我们基于 DQP 的模型在脑电图语言数据集上采用 "只留一个被试"（LOSO）交叉验证策略，分类准确率达到了 95.68%。此外，DQP-LangNet 还采用了可解释特征工程（XFE）结构，以获得特定信道的可解释结果。我们提出的 DQP-LangNet 模型可用于神经科学领域的其他应用。

{"title":"Automated EEG-based language detection using directed quantum pattern technique","authors":"","doi":"10.1016/j.asoc.2024.112301","DOIUrl":"10.1016/j.asoc.2024.112301","url":null,"abstract":"<div><div>Electroencephalogram (EEG) signals contain complex useful information about brain activities. These EEG signals are noisy, highly varying and nonstationary in nature. Hence, extracting meaningful information from these signals is challenging. The existing machine learning systems struggle to capture the minute changes from the signals and yield high performance.</div><div>This study introduces a novel quantum-inspired feature extraction technique called Directed Quantum Pattern (DQP), designed to address these challenges by using a lattice structure to capture directional binary features. These directions (paths) are computed using a maximum function providing a dynamic and adaptive feature representation.</div><div>This paper presents a novel DQP-LangNet developed using DQP for automated classification of two- languages using EEG signals. We have proposed a hybrid approach, combining DQP, statistical features, and multi-level discrete wavelet transform (MDWT) to extract salient features similar to the deep learning approach. The EEG dataset consisting of 14 channels, produces 7 feature vectors per channel, yielding 98 feature vectors. Neighborhood component analysis and Chi-square (Chi2) feature selection approaches generated 196 feature vectors.</div><div>In addition to the innovative feature extraction a new classification structure called “t” is proposed k-nearest neighbor (tkNN) and support vector machine (tSVM) classifiers are employed. Using the proposed tkNN and tSVM classifiers, 392 (=196×2) classifier-based outcomes are obtained. To further improve classification performance, we applied the iterative majority voting (IMV) technique to automatically select the best result.</div><div>Our DQP-based model achieved a classification accuracy of 95.68 %using EEG language dataset with leave-one-subject-out (LOSO) cross-validation strategy. Also, an explainable feature engineering (XFE) structure of DQP-LangNet is employed to obtain channel-specific explainable results. Our proposed DQP-LangNet model can be employed for other applications in neuroscience.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":null,"pages":null},"PeriodicalIF":7.2,"publicationDate":"2024-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142427659","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}

引用次数: 0

Unravelling sleep patterns: Supervised contrastive learning with self-attention for sleep stage classification 揭示睡眠模式：利用自我关注的监督对比学习进行睡眠阶段分类

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2024-10-05 DOI: 10.1016/j.asoc.2024.112298

Sleep data scoring is a crucial and primary step for diagnosing sleep disorders to know the sleep stages from the PSG signals. This study uses supervised contrastive learning with a self-attention mechanism to classify sleep stages. We propose a deep learning framework for automatic sleep stage classification, which involves two training phases: (1) the feature representation learning phase, in which the feature representation network (encoder) learns to extract features from the electroencephalogram (EEG) signals, and (2) the classification network training phase, where a pre-trained encoder (trained during phase I) along with the classifier head is fine-tuned for the classification task. The PSG data shows a non-uniform distribution of sleep stages, with wake (W) (around 30% of total samples) and N2 stages (around 58% and 37% of total samples in Physionet EDF-Sleep 2013 and 2018 datasets, respectively) being more prevalent, leading to an imbalanced dataset. The imbalanced data issue is addressed using a weighted softmax cross-entropy loss function that assigns higher weights to minority sleep stages. Additionally, an oversampling technique (the synthetic minority oversampling technique (SMOTE) (Chawla et al., 2002)[1] ) is applied to generate synthetic samples for minority classes. The proposed model is evaluated on the Physionet EDF-Sleep 2013 and 2018 datasets using Fpz-Cz and Pz-Oz EEG channels. It achieved an overall accuracy of 94.1%, a macro F1 score of 92.64, and a Cohen’s Kappa coefficient of 0.92. Ablation studies demonstrated the importance of triplet loss-based pre-training and oversampling for enhancing performance. The proposed model requires minimal pre-processing, eliminating the need for extensive signal processing expertise, and thus is well-suited for clinicians diagnosing sleep disorders.

从 PSG 信号中了解睡眠阶段是诊断睡眠障碍的关键和首要步骤。本研究利用具有自我注意机制的监督对比学习来对睡眠阶段进行分类。我们提出了一种用于自动进行睡眠阶段分类的深度学习框架，它包括两个训练阶段：（1）特征表示学习阶段，在这一阶段中，特征表示网络（编码器）学习从脑电信号中提取特征；（2）分类网络训练阶段，在这一阶段中，预先训练好的编码器（在第一阶段中训练好）与分类器头一起针对分类任务进行微调。PSG 数据显示睡眠阶段分布不均匀，唤醒（W）阶段（约占总样本的 30%）和 N2 阶段（在 Physionet EDF-Sleep 2013 和 2018 数据集中分别约占总样本的 58% 和 37%）更为普遍，导致数据集不平衡。解决数据不平衡问题的方法是使用加权软最大交叉熵损失函数，为少数睡眠阶段分配更高的权重。此外，还采用了超采样技术（合成少数群体超采样技术（SMOTE）（Chawla 等人，2002 年）[1] ）为少数群体类别生成合成样本。利用 Fpz-Cz 和 Pz-Oz 脑电图通道，在 Physionet EDF-Sleep 2013 和 2018 数据集上对所提出的模型进行了评估。其总体准确率达到 94.1%，宏观 F1 得分为 92.64，科恩卡帕系数为 0.92。消融研究表明，基于三重损失的预训练和超采样对提高性能非常重要。所提出的模型只需极少的预处理，无需大量的信号处理专业知识，因此非常适合临床医生诊断睡眠障碍。

{"title":"Unravelling sleep patterns: Supervised contrastive learning with self-attention for sleep stage classification","authors":"","doi":"10.1016/j.asoc.2024.112298","DOIUrl":"10.1016/j.asoc.2024.112298","url":null,"abstract":"<div><div>Sleep data scoring is a crucial and primary step for diagnosing sleep disorders to know the sleep stages from the PSG signals. This study uses supervised contrastive learning with a self-attention mechanism to classify sleep stages. We propose a deep learning framework for automatic sleep stage classification, which involves two training phases: (1) the feature representation learning phase, in which the feature representation network (encoder) learns to extract features from the electroencephalogram (EEG) signals, and (2) the classification network training phase, where a pre-trained encoder (trained during phase I) along with the classifier head is fine-tuned for the classification task. The PSG data shows a non-uniform distribution of sleep stages, with wake (W) (around 30% of total samples) and N2 stages (around 58% and 37% of total samples in Physionet EDF-Sleep 2013 and 2018 datasets, respectively) being more prevalent, leading to an imbalanced dataset. The imbalanced data issue is addressed using a weighted softmax cross-entropy loss function that assigns higher weights to minority sleep stages. Additionally, an oversampling technique (the synthetic minority oversampling technique (SMOTE) (Chawla et al., 2002)<span><span>[1]</span></span> ) is applied to generate synthetic samples for minority classes. The proposed model is evaluated on the Physionet EDF-Sleep 2013 and 2018 datasets using Fpz-Cz and Pz-Oz EEG channels. It achieved an overall accuracy of 94.1%, a macro F1 score of 92.64, and a Cohen’s Kappa coefficient of 0.92. Ablation studies demonstrated the importance of triplet loss-based pre-training and oversampling for enhancing performance. The proposed model requires minimal pre-processing, eliminating the need for extensive signal processing expertise, and thus is well-suited for clinicians diagnosing sleep disorders.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":null,"pages":null},"PeriodicalIF":7.2,"publicationDate":"2024-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142441287","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}

引用次数: 0

Enhancing driver attention and road safety through EEG-informed deep reinforcement learning and soft computing 通过基于脑电图的深度强化学习和软计算提高驾驶员注意力和道路安全

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2024-10-05 DOI: 10.1016/j.asoc.2024.112320

This paper introduces a transformative edge computing-based approach for enhancing driver attention and road safety using EEG-driven deep reinforcement learning (DRL). As driver inattention remains a significant factor in accidents, real-time cognitive state monitoring enabled by in-vehicle edge devices offers new promise. Our method leverages EEG data collected from drivers using headsets, analyzing signals related to visual attention. Edge computing resources in the vehicle extract features and classify attention levels using Deep Q-Network (DQN) and Proximal Policy Optimization (PPO) models trained to approximate optimal driving decisions. A novel reward structure combining driving performance and attention guides the models. Our edge computing-powered framework reacts within critical time latencies to maximize attention through interventions adapting to the driving environment. Results demonstrate the effectiveness of this approach, with PPO agent on edge devices achieving high average rewards up to 489,752.4 and 99.3% reward as accuracy in classifying attention states, thereby significantly outperforming traditional methods. This underscores edge computing’s potential to enable real-time integration of neuroscience and AI, advancing road safety. The edge resources deliver time-critical analysis and adaptation, while connectivity to the fog and cloud allows optimizing and learning at scale across populations. This research pioneers a new epoch for road safety powered by edge intelligence.

本文介绍了一种基于边缘计算的变革性方法，利用脑电图驱动的深度强化学习（DRL）提高驾驶员注意力和道路安全。由于驾驶员注意力不集中仍然是导致交通事故的一个重要因素，由车载边缘设备实现的实时认知状态监测带来了新的希望。我们的方法利用从使用耳机的驾驶员处收集的脑电图数据，分析与视觉注意力相关的信号。车载边缘计算资源使用深度 Q 网络（DQN）和近端策略优化（PPO）模型提取特征并对注意力水平进行分类，这些模型经过训练，可近似地做出最佳驾驶决策。结合驾驶性能和注意力的新型奖励结构为模型提供指导。我们的边缘计算驱动框架能在关键的时间延迟内做出反应，通过适应驾驶环境的干预措施最大限度地提高注意力。结果证明了这种方法的有效性，边缘设备上的 PPO 代理在注意力状态分类方面实现了高达 489,752.4 的高平均奖励和 99.3% 的奖励准确率，从而大大优于传统方法。这凸显了边缘计算在实现神经科学与人工智能的实时整合、促进道路安全方面的潜力。边缘资源可提供时间关键型分析和适应，而与雾和云的连接则可实现跨人群的大规模优化和学习。这项研究开创了由边缘智能驱动的道路安全新纪元。

{"title":"Enhancing driver attention and road safety through EEG-informed deep reinforcement learning and soft computing","authors":"","doi":"10.1016/j.asoc.2024.112320","DOIUrl":"10.1016/j.asoc.2024.112320","url":null,"abstract":"<div><div>This paper introduces a transformative edge computing-based approach for enhancing driver attention and road safety using EEG-driven deep reinforcement learning (DRL). As driver inattention remains a significant factor in accidents, real-time cognitive state monitoring enabled by in-vehicle edge devices offers new promise. Our method leverages EEG data collected from drivers using headsets, analyzing signals related to visual attention. Edge computing resources in the vehicle extract features and classify attention levels using Deep Q-Network (DQN) and Proximal Policy Optimization (PPO) models trained to approximate optimal driving decisions. A novel reward structure combining driving performance and attention guides the models. Our edge computing-powered framework reacts within critical time latencies to maximize attention through interventions adapting to the driving environment. Results demonstrate the effectiveness of this approach, with PPO agent on edge devices achieving high average rewards up to 489,752.4 and 99.3% reward as accuracy in classifying attention states, thereby significantly outperforming traditional methods. This underscores edge computing’s potential to enable real-time integration of neuroscience and AI, advancing road safety. The edge resources deliver time-critical analysis and adaptation, while connectivity to the fog and cloud allows optimizing and learning at scale across populations. This research pioneers a new epoch for road safety powered by edge intelligence.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":null,"pages":null},"PeriodicalIF":7.2,"publicationDate":"2024-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142527754","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"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

Adaptive fuzzy coordinated control design for wind turbine using gray wolf optimization algorithm 使用灰狼优化算法的风力涡轮机自适应模糊协调控制设计

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2024-10-05 DOI: 10.1016/j.asoc.2024.112319

Due to the randomness and intermittency of wind speed, the actual output power curve of a wind turbine (WT) deviates greatly from the theoretical power curve, thereby reducing the power generation capacity of the WT. An adaptive fuzzy coordinated control (AFCC) of WT is presented in this study to improve the power generation of WT. Firstly, a multi-objective optimization model (MOOM) for WT output power, generator speed and pitch angle is established, and its optimal solution set is used as the input eigenvector of a novel effective wind speed soft sensor (NEWSSS) model, which is modeled with kernel extreme learning machine (KELM). Secondly, a novel improved gray wolf optimization (NIGWO) algorithm is presented by improving the convergence factor and adaptive weights, which is used to solve MOOM and optimize the parameters of KELM. A variable pitch control (VPC) is designed by estimating the effective wind speed. Finally, an adaptive fuzzy control (AFC) is presented for WT. Based on the AFC and VPC, an AFCC for pitch angle and generator torque is designed for WT. The high measuring precision of NEWSSS and the good robustness and dynamic performance of AFCC are demonstrated by the simulation results.

由于风速的随机性和间歇性，风力涡轮机（WT）的实际输出功率曲线与理论功率曲线偏差很大，从而降低了 WT 的发电能力。本研究提出了一种风力涡轮机自适应模糊协调控制（AFCC），以提高风力涡轮机的发电量。首先，建立了风电机组输出功率、发电机转速和变桨角的多目标优化模型（MOOM），并将其最优解集作为新型有效风速软传感器（NEWSSS）模型的输入特征向量，该模型采用内核极端学习机（KELM）建模。其次，通过改进收敛因子和自适应权重，提出了一种新型改进灰狼优化（NIGWO）算法，用于求解 MOOM 和优化 KELM 的参数。通过估计有效风速，设计了变桨距控制（VPC）。最后，提出了针对 WT 的自适应模糊控制（AFC）。在 AFC 和 VPC 的基础上，为 WT 设计了变桨角和发电机转矩的 AFCC。仿真结果表明，NEWSSS 测量精度高，AFCC 具有良好的鲁棒性和动态性能。

{"title":"Adaptive fuzzy coordinated control design for wind turbine using gray wolf optimization algorithm","authors":"","doi":"10.1016/j.asoc.2024.112319","DOIUrl":"10.1016/j.asoc.2024.112319","url":null,"abstract":"<div><div>Due to the randomness and intermittency of wind speed, the actual output power curve of a wind turbine (WT) deviates greatly from the theoretical power curve, thereby reducing the power generation capacity of the WT. An adaptive fuzzy coordinated control (AFCC) of WT is presented in this study to improve the power generation of WT. Firstly, a multi-objective optimization model (MOOM) for WT output power, generator speed and pitch angle is established, and its optimal solution set is used as the input eigenvector of a novel effective wind speed soft sensor (NEWSSS) model, which is modeled with kernel extreme learning machine (KELM). Secondly, a novel improved gray wolf optimization (NIGWO) algorithm is presented by improving the convergence factor and adaptive weights, which is used to solve MOOM and optimize the parameters of KELM. A variable pitch control (VPC) is designed by estimating the effective wind speed. Finally, an adaptive fuzzy control (AFC) is presented for WT. Based on the AFC and VPC, an AFCC for pitch angle and generator torque is designed for WT. The high measuring precision of NEWSSS and the good robustness and dynamic performance of AFCC are demonstrated by the simulation results.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":null,"pages":null},"PeriodicalIF":7.2,"publicationDate":"2024-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142427663","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}

引用次数: 0

Novel imbalanced multi-class fault diagnosis method using transfer learning and oversampling strategies-based multi-layer support vector machines (ML-SVMs) 使用基于迁移学习和超采样策略的多层支持向量机 (ML-SVM) 的新型不平衡多类故障诊断方法

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing

Pub Date : 2024-10-05 DOI: 10.1016/j.asoc.2024.112324

For health monitoring and fault diagnosis of critical mechanical system components, historical data related to equipment failures are often limited and exhibit varying imbalanced multi-class characteristics (e.g., with noisy and time-series data). Moreover, fault diagnosis frameworks based on traditional resampling algorithms (e.g., SMOTE) mostly heavily rely on manual feature extraction, making them difficult to adapt to diverse working conditions or objects. To address these challenges, we propose a novel end-to-end imbalanced multi-class fault diagnosis architecture using transfer learning and oversampling strategies-based multi-layer support vector machines (ML-SVMs). ML-SVMs utilize a VGG-based deep migration feature extraction method to extract features from original time-domain vibration signals, employing natural source domain weights to reduce dependence on human experience and sample size. Then, ML-SVMs introduce ISCOTE (i.e., the first and second layers of ML-SVMs), an improved version of the sample-characteristic over-sampling technique (SCOTE). ISCOTE generates more effective and reasonable samples for each fault class through a scaling factor and iterative optimization mechanism, whether in noisy feature spaces with fuzzy boundaries or in clear boundary feature spaces. Finally, in the third layer of ML-SVMs, multi-class SVMs (e.g., LS-SVMs and standard SVMs) are utilized to train balanced feature samples and derive classification models with strong generalization ability. The effectiveness of ML-SVMs is demonstrated through 16 fault diagnosis instances using CWRU and IMS bearing data, PHM 2010 and TTWD tool wear data. Results indicate that ML-SVMs outperform 8 well-known oversampling-based algorithms in fault diagnosis recognition rates and algorithm robustness. It has offered a feasible architecture for multi-class imbalanced fault scenarios with limited data and multiple adverse features.

对于关键机械系统组件的健康监测和故障诊断而言，与设备故障相关的历史数据往往有限，且呈现出不同的不平衡多类特征（如噪声数据和时间序列数据）。此外，基于传统重采样算法的故障诊断框架（如 SMOTE）在很大程度上依赖于人工特征提取，很难适应不同的工作条件或对象。为了应对这些挑战，我们提出了一种新颖的端到端不平衡多类故障诊断架构，该架构采用基于迁移学习和超采样策略的多层支持向量机（ML-SVM）。ML-SVM 利用基于 VGG 的深度迁移特征提取方法，从原始时域振动信号中提取特征，并采用自然源域权重以减少对人类经验和样本大小的依赖。然后，ML-SVM 引入了 ISCOTE（即 ML-SVM 的第一层和第二层），它是样本特征超采样技术（SCOTE）的改进版本。无论是在边界模糊的噪声特征空间，还是在边界清晰的特征空间，ISCOTE 都能通过缩放因子和迭代优化机制为每个故障类别生成更有效、更合理的样本。最后，在 ML-SVM 的第三层，利用多类 SVM（如 LS-SVM 和标准 SVM）来训练均衡的特征样本，并得出具有较强泛化能力的分类模型。利用 CWRU 和 IMS 轴承数据、PHM 2010 和 TTWD 工具磨损数据，通过 16 个故障诊断实例证明了 ML-SVM 的有效性。结果表明，ML-SVM 在故障诊断识别率和算法鲁棒性方面优于 8 种著名的基于超采样的算法。它为具有有限数据和多种不利特征的多类不平衡故障场景提供了一种可行的架构。

{"title":"Novel imbalanced multi-class fault diagnosis method using transfer learning and oversampling strategies-based multi-layer support vector machines (ML-SVMs)","authors":"","doi":"10.1016/j.asoc.2024.112324","DOIUrl":"10.1016/j.asoc.2024.112324","url":null,"abstract":"<div><div>For health monitoring and fault diagnosis of critical mechanical system components, historical data related to equipment failures are often limited and exhibit varying imbalanced multi-class characteristics (e.g., with noisy and time-series data). Moreover, fault diagnosis frameworks based on traditional resampling algorithms (e.g., SMOTE) mostly heavily rely on manual feature extraction, making them difficult to adapt to diverse working conditions or objects. To address these challenges, we propose a novel end-to-end imbalanced multi-class fault diagnosis architecture using transfer learning and oversampling strategies-based multi-layer support vector machines (ML-SVMs). ML-SVMs utilize a VGG-based deep migration feature extraction method to extract features from original time-domain vibration signals, employing natural source domain weights to reduce dependence on human experience and sample size. Then, ML-SVMs introduce ISCOTE (i.e., the first and second layers of ML-SVMs), an improved version of the sample-characteristic over-sampling technique (SCOTE). ISCOTE generates more effective and reasonable samples for each fault class through a scaling factor and iterative optimization mechanism, whether in noisy feature spaces with fuzzy boundaries or in clear boundary feature spaces. Finally, in the third layer of ML-SVMs, multi-class SVMs (e.g., LS-SVMs and standard SVMs) are utilized to train balanced feature samples and derive classification models with strong generalization ability. The effectiveness of ML-SVMs is demonstrated through 16 fault diagnosis instances using CWRU and IMS bearing data, PHM 2010 and TTWD tool wear data. Results indicate that ML-SVMs outperform 8 well-known oversampling-based algorithms in fault diagnosis recognition rates and algorithm robustness. It has offered a feasible architecture for multi-class imbalanced fault scenarios with limited data and multiple adverse features.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":null,"pages":null},"PeriodicalIF":7.2,"publicationDate":"2024-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142427826","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}

引用次数: 0