Knowledge-Based Systems最新文献

英文中文

Predicting restaurant survival using nationwide Google Maps data 利用全国谷歌地图数据预测餐厅生存状况

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

Knowledge-Based Systems

Pub Date : 2025-02-21 DOI: 10.1016/j.knosys.2025.113198

Tomasz Starakiewicz, Piotr Wójcik

The restaurant sector is pivotal to firm exit research, which influences economic policy and managerial strategy recommendations. Recent studies using online data are based on geographically limited datasets and have largely omitted temporal dynamics in user interactions. Additionally, these studies rely on manual labeling for text analysis, a resource-intensive approach. Built upon the case of Poland, our study introduces the first comprehensive, nationwide analysis of restaurant survival using Google Maps data. We enhance predictive model performance by incorporating time-sensitive user interactions. Our model controls for established determinants of business exit and proves robust regarding data quality issues associated with user-provided business directories. We apply an efficient, label-free method for extracting semantic content from reviews, thereby creating useful features for firm exit prediction. Furthermore, we present an efficient feature selection strategy using hierarchical agglomerative clustering that retains predictive power while reducing the model complexity. Our model has broad applications ranging from credit scoring to early-warning systems for business closures, while our data collection method opens doors to large-scale firm exit studies in regions where official records are lacking and online sources used in previous studies are less prevalent.

{"title":"Predicting restaurant survival using nationwide Google Maps data","authors":"Tomasz Starakiewicz, Piotr Wójcik","doi":"10.1016/j.knosys.2025.113198","DOIUrl":"10.1016/j.knosys.2025.113198","url":null,"abstract":"<div><div>The restaurant sector is pivotal to firm exit research, which influences economic policy and managerial strategy recommendations. Recent studies using online data are based on geographically limited datasets and have largely omitted temporal dynamics in user interactions. Additionally, these studies rely on manual labeling for text analysis, a resource-intensive approach. Built upon the case of Poland, our study introduces the first comprehensive, nationwide analysis of restaurant survival using Google Maps data. We enhance predictive model performance by incorporating time-sensitive user interactions. Our model controls for established determinants of business exit and proves robust regarding data quality issues associated with user-provided business directories. We apply an efficient, label-free method for extracting semantic content from reviews, thereby creating useful features for firm exit prediction. Furthermore, we present an efficient feature selection strategy using hierarchical agglomerative clustering that retains predictive power while reducing the model complexity. Our model has broad applications ranging from credit scoring to early-warning systems for business closures, while our data collection method opens doors to large-scale firm exit studies in regions where official records are lacking and online sources used in previous studies are less prevalent.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"314 ","pages":"Article 113198"},"PeriodicalIF":7.2,"publicationDate":"2025-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143534357","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

DHR-BLS: A Huber-type robust broad learning system with its distributed version DHR-BLS：胡贝尔式稳健广义学习系统及其分布式版本

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

Knowledge-Based Systems

Pub Date : 2025-02-21 DOI: 10.1016/j.knosys.2025.113184

Yuao Zhang , Shuya Ke , Jing Li , Weihua Liu , Jueliang Hu , Kaixiang Yang

The broad learning system (BLS) is a recently developed neural network framework recognized for its efficiency and effectiveness in handling high-dimensional data with a flat network architecture. However, traditional BLS models are highly sensitive to outliers and noisy data, which can significantly degrade performance. While incorporating the

ℓ_{1}

-norm loss function enhances robustness against outliers, it often compromises performance on clean datasets. To address this limitation, we propose the Huber-type robust broad learning system (HR-BLS), which integrates the Huber loss function into BLS, effectively combining the strengths of both

ℓ_{1}

-norm and

ℓ_{2}

-norm loss functions to achieve balanced robustness against data anomalies. Moreover, the elastic-net regularization is included to simultaneously enhance model stability and promote sparsity. To effectively manage large-scale and distributed data, we extend HR-BLS by introducing the distributed Huber-type robust broad learning system (DHR-BLS). Given the non-differentiability of the

ℓ_{1}

-norm, traditional gradient-based optimization methods are insufficient. Therefore, we adopt the alternating direction method of multipliers (ADMM) to train, ensuring convergence through the use of appropriate constraints. Experimental results on both synthetic and benchmark datasets show that HR-BLS outperforms traditional BLS and other state-of-the-art robust learning methods in terms of accuracy and robustness. Furthermore, DHR-BLS demonstrates exceptional scalability and effectiveness, making it suitable for distributed learning environments.

{"title":"DHR-BLS: A Huber-type robust broad learning system with its distributed version","authors":"Yuao Zhang , Shuya Ke , Jing Li , Weihua Liu , Jueliang Hu , Kaixiang Yang","doi":"10.1016/j.knosys.2025.113184","DOIUrl":"10.1016/j.knosys.2025.113184","url":null,"abstract":"<div><div>The broad learning system (BLS) is a recently developed neural network framework recognized for its efficiency and effectiveness in handling high-dimensional data with a flat network architecture. However, traditional BLS models are highly sensitive to outliers and noisy data, which can significantly degrade performance. While incorporating the <span><math><msub><mrow><mi>ℓ</mi></mrow><mrow><mn>1</mn></mrow></msub></math></span>-norm loss function enhances robustness against outliers, it often compromises performance on clean datasets. To address this limitation, we propose the Huber-type robust broad learning system (HR-BLS), which integrates the Huber loss function into BLS, effectively combining the strengths of both <span><math><msub><mrow><mi>ℓ</mi></mrow><mrow><mn>1</mn></mrow></msub></math></span>-norm and <span><math><msub><mrow><mi>ℓ</mi></mrow><mrow><mn>2</mn></mrow></msub></math></span>-norm loss functions to achieve balanced robustness against data anomalies. Moreover, the elastic-net regularization is included to simultaneously enhance model stability and promote sparsity. To effectively manage large-scale and distributed data, we extend HR-BLS by introducing the distributed Huber-type robust broad learning system (DHR-BLS). Given the non-differentiability of the <span><math><msub><mrow><mi>ℓ</mi></mrow><mrow><mn>1</mn></mrow></msub></math></span>-norm, traditional gradient-based optimization methods are insufficient. Therefore, we adopt the alternating direction method of multipliers (ADMM) to train, ensuring convergence through the use of appropriate constraints. Experimental results on both synthetic and benchmark datasets show that HR-BLS outperforms traditional BLS and other state-of-the-art robust learning methods in terms of accuracy and robustness. Furthermore, DHR-BLS demonstrates exceptional scalability and effectiveness, making it suitable for distributed learning environments.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"314 ","pages":"Article 113184"},"PeriodicalIF":7.2,"publicationDate":"2025-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143471538","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

Intermediate triple table: A general architecture for virtual knowledge graphs

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

Knowledge-Based Systems

Pub Date : 2025-02-21 DOI: 10.1016/j.knosys.2025.113179

Julián Arenas-Guerrero, Oscar Corcho, María S. Pérez

Virtual knowledge graphs (VKGs) have been widely applied to access relational data with a semantic layer by using an ontology in use cases that are dynamic in nature. However, current VKG techniques focus mainly on accessing a single relational database and remain largely unstudied for data integration with several heterogeneous data sources. To overcome this limitation, we propose intermediate triple table (ITT), a general VKG architecture to access multiple and diverse data sources. Our proposal is based on data shipping and addresses heterogeneity by adopting a schema-oblivious graph representation that intervenes between the sources and the queries. We minimize data computation by just materializing a relevant subgraph for a specific query. We employ star-shaped query processing and extend this technique to mapping candidate selection. For rapid materialization of the ITT, we apply a mapping partitioning technique to parallelize mapping execution, which also guarantees duplicate-free subgraphs and reduces memory consumption. We use SPARQL-to-SQL query translation to homogeneously evaluate queries over the ITT and execute them with an in-process analytical store. We implemented ITT on top of a knowledge graph materialization engine and evaluated it with two VKG benchmarks. The experimental results show that our proposal outperforms state-of-the-art techniques for complex graph queries in terms of execution time. It also decreases the number of timeouts although it uses more memory as a trade-off. The experiments also demonstrate the source independence of the architecture on a mixed distribution of data with SQL and document stores together with various file formats.

{"title":"Intermediate triple table: A general architecture for virtual knowledge graphs","authors":"Julián Arenas-Guerrero, Oscar Corcho, María S. Pérez","doi":"10.1016/j.knosys.2025.113179","DOIUrl":"10.1016/j.knosys.2025.113179","url":null,"abstract":"<div><div>Virtual knowledge graphs (VKGs) have been widely applied to access relational data with a semantic layer by using an ontology in use cases that are dynamic in nature. However, current VKG techniques focus mainly on accessing a single relational database and remain largely unstudied for data integration with several heterogeneous data sources. To overcome this limitation, we propose <em>intermediate triple table</em> (<span>ITT</span>), a general VKG architecture to access multiple and diverse data sources. Our proposal is based on data shipping and addresses heterogeneity by adopting a schema-oblivious graph representation that intervenes between the sources and the queries. We minimize data computation by just materializing a relevant subgraph for a specific query. We employ star-shaped query processing and extend this technique to mapping candidate selection. For rapid materialization of the <span>ITT</span>, we apply a mapping partitioning technique to parallelize mapping execution, which also guarantees duplicate-free subgraphs and reduces memory consumption. We use SPARQL-to-SQL query translation to homogeneously evaluate queries over the <span>ITT</span> and execute them with an in-process analytical store. We implemented <span>ITT</span> on top of a knowledge graph materialization engine and evaluated it with two VKG benchmarks. The experimental results show that our proposal outperforms state-of-the-art techniques for complex graph queries in terms of execution time. It also decreases the number of timeouts although it uses more memory as a trade-off. The experiments also demonstrate the source independence of the architecture on a mixed distribution of data with SQL and document stores together with various file formats.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"314 ","pages":"Article 113179"},"PeriodicalIF":7.2,"publicationDate":"2025-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143480692","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

An enhanced BiGAN architecture for network intrusion detection

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

Knowledge-Based Systems

Pub Date : 2025-02-21 DOI: 10.1016/j.knosys.2025.113178

Mohammad Arafah , Iain Phillips , Asma Adnane , Mohammad Alauthman , Nauman Aslam

Intrusion detection systems face significant challenges in handling high-dimensional, large-scale, and imbalanced network traffic data. This paper proposes a new architecture combining a denoising autoencoder (AE) and a Wasserstein Generative Adversarial Network (WGAN) to address these challenges. The AE-WGAN model extracts high-representative features and generates realistic synthetic attacks, effectively resolving data imbalance and enhancing anomaly-based intrusion detection. Our extensive experiments on NSL-KDD and CICIDS-2017 datasets demonstrate superior performance, achieving 98% accuracy and 99% F1-score in binary classification, surpassing recent approaches by 7%–15%. In multiclass cases, the model achieves 89% precision for DoS attacks and 84% for Probe attacks, while maintaining 79% precision for rare U2R attacks. Time complexity analysis reveals 23% reduced training time while maintaining high-quality synthetic attack generation, contributing a robust framework capable of handling modern network traffic complexities and evolving cyber threats.

引用次数: 0

Building robust deep recommender systems: Utilizing a weighted adversarial noise propagation framework with robust fine-tuning modules

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

Knowledge-Based Systems

Pub Date : 2025-02-20 DOI: 10.1016/j.knosys.2025.113181

Fulan Qian, Wenbin Chen, Hai Chen, Jinggang Liu, Shu Zhao, Yanping Zhang

The performance of deep recommendation algorithms decreases significantly under adversarial attacks. While some approaches improve the recommender system robustness via adversarial training, they primarily target shallow models or rely on coarse-grained noise, so that deep models remain vulnerable. This study proposes a new adversarial training framework, the Random Adversarial Weight Perturbation Framework Equipped with Robust Fine-Tuning (RAWP-FT). Specifically, RAWP-FT first performs adversarial training of deep models by introducing more fine-grained adversarial noise into the hidden layer weight parameters. Subsequently, RAWP-FT identifies and targets the modules or layers with the lowest robustness after adversarial training and performs specialized adversarial training and fine-tuning to improve the model robustness further. Experiments demonstrate that RAWP-FT significantly enhances the robustness of deep recommendation models. We apply RAWP-FT to MLP and other deep models, highlighting its ability to strengthen vulnerable components through robust critical fine-tuning. Experiments on four publicly available datasets confirm that RAWP-FT-trained models can withstand adversarial noise while maintaining performance.

{"title":"Building robust deep recommender systems: Utilizing a weighted adversarial noise propagation framework with robust fine-tuning modules","authors":"Fulan Qian, Wenbin Chen, Hai Chen, Jinggang Liu, Shu Zhao, Yanping Zhang","doi":"10.1016/j.knosys.2025.113181","DOIUrl":"10.1016/j.knosys.2025.113181","url":null,"abstract":"<div><div>The performance of deep recommendation algorithms decreases significantly under adversarial attacks. While some approaches improve the recommender system robustness via adversarial training, they primarily target shallow models or rely on coarse-grained noise, so that deep models remain vulnerable. This study proposes a new adversarial training framework, the Random Adversarial Weight Perturbation Framework Equipped with Robust Fine-Tuning (RAWP-FT). Specifically, RAWP-FT first performs adversarial training of deep models by introducing more fine-grained adversarial noise into the hidden layer weight parameters. Subsequently, RAWP-FT identifies and targets the modules or layers with the lowest robustness after adversarial training and performs specialized adversarial training and fine-tuning to improve the model robustness further. Experiments demonstrate that RAWP-FT significantly enhances the robustness of deep recommendation models. We apply RAWP-FT to MLP and other deep models, highlighting its ability to strengthen vulnerable components through robust critical fine-tuning. Experiments on four publicly available datasets confirm that RAWP-FT-trained models can withstand adversarial noise while maintaining performance.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"314 ","pages":"Article 113181"},"PeriodicalIF":7.2,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143509731","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

SAM-LAD: Segment Anything Model meets zero-shot logic anomaly detection

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

Knowledge-Based Systems

Pub Date : 2025-02-20 DOI: 10.1016/j.knosys.2025.113176

Yun Peng, Xiao Lin, Nachuan Ma, Jiayuan Du, Chuangwei Liu, Chengju Liu, Qijun Chen

Visual anomaly detection is vital in real-world applications, such as industrial defect detection and medical diagnosis. However, most existing methods focus on local structural anomalies and fail to detect higher-level functional anomalies under logical conditions. Although recent studies have explored logical anomaly detection, they can only address simple anomalies like missing or addition and show poor generalizability due to being heavily data-driven. To fill this gap, we propose SAM-LAD, a zero-shot, plug-and-play framework for anomaly detection in any scene. First, we obtain a query image’s feature map using a pre-trained backbone. Simultaneously, we retrieve the reference images and their corresponding feature maps via the nearest neighbor search. Then, we introduce the Segment Anything Model (SAM) to obtain object masks of the query and reference images. Each object mask is multiplied by the entire image’s feature map to obtain object feature maps. Next, an Object Matching Model (OMM) is proposed to match objects in the query and reference images. To facilitate object matching, we propose a Dynamic Channel Graph Attention (DCGA) module, treating each object as a keypoint and converting its feature maps into feature vectors. Finally, based on the object matching relations, an Anomaly Measurement Model (AMM) is proposed to detect objects with logical anomalies. Structural anomalies in the objects can also be detected. We validate our proposed SAM-LAD using various benchmarks, including industrial datasets (MVTec Loco AD, MVTec AD), and the logical dataset (DigitAnatomy). Extensive experimental results demonstrate that SAM-LAD outperforms existing SoTA methods, particularly in detecting logical anomalies.

{"title":"SAM-LAD: Segment Anything Model meets zero-shot logic anomaly detection","authors":"Yun Peng, Xiao Lin, Nachuan Ma, Jiayuan Du, Chuangwei Liu, Chengju Liu, Qijun Chen","doi":"10.1016/j.knosys.2025.113176","DOIUrl":"10.1016/j.knosys.2025.113176","url":null,"abstract":"<div><div>Visual anomaly detection is vital in real-world applications, such as industrial defect detection and medical diagnosis. However, most existing methods focus on local structural anomalies and fail to detect higher-level functional anomalies under logical conditions. Although recent studies have explored logical anomaly detection, they can only address simple anomalies like missing or addition and show poor generalizability due to being heavily data-driven. To fill this gap, we propose SAM-LAD, a zero-shot, plug-and-play framework for anomaly detection in any scene. First, we obtain a query image’s feature map using a pre-trained backbone. Simultaneously, we retrieve the reference images and their corresponding feature maps via the nearest neighbor search. Then, we introduce the Segment Anything Model (SAM) to obtain object masks of the query and reference images. Each object mask is multiplied by the entire image’s feature map to obtain object feature maps. Next, an Object Matching Model (OMM) is proposed to match objects in the query and reference images. To facilitate object matching, we propose a Dynamic Channel Graph Attention (DCGA) module, treating each object as a keypoint and converting its feature maps into feature vectors. Finally, based on the object matching relations, an Anomaly Measurement Model (AMM) is proposed to detect objects with logical anomalies. Structural anomalies in the objects can also be detected. We validate our proposed SAM-LAD using various benchmarks, including industrial datasets (MVTec Loco AD, MVTec AD), and the logical dataset (DigitAnatomy). Extensive experimental results demonstrate that SAM-LAD outperforms existing SoTA methods, particularly in detecting logical anomalies.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"314 ","pages":"Article 113176"},"PeriodicalIF":7.2,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143511341","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

HCN-RLR-CAN: A novel human-computer negotiation model based on round-level recurrence and causal attention networks

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

Knowledge-Based Systems

Pub Date : 2025-02-20 DOI: 10.1016/j.knosys.2025.113180

Jianting Zhang , Xudong Luo , Xiaojun Xie

Recent advances in Human-Computer Negotiation (HCN) systems have shown promising progress in simulating complex negotiation dialogues. However, these systems still need to grapple with critical challenges such as limited adaptability to proper negotiation tones, ineffective capture of long-term dependencies, and constraints in generating diverse, natural, and strategically sound responses. To address these limitations, we propose the HCN-RLR-CAN, a novel HCN model based on Round-Level Recurrence (RLR) and Causal Attention Network (CAN). Our approach, which uniquely processes dialogues as role-based parallel texts and employs a hierarchical encoder to comprehend user dialogue history, offers a fresh perspective on addressing these challenges. The model employs causal attention learning modules to separately model linguistic strategies and dialogue acts. Additionally, it employs a round-level recursive decoding mechanism that generates responses by synthesising historical dialogue information, dialogue act and strategy encodings, and previous decoding states. With its significant performance advantages over baseline models, the HCN-RLR-CAN model has the potential to inspire a new wave of research and development in the field of HCN systems.

{"title":"HCN-RLR-CAN: A novel human-computer negotiation model based on round-level recurrence and causal attention networks","authors":"Jianting Zhang , Xudong Luo , Xiaojun Xie","doi":"10.1016/j.knosys.2025.113180","DOIUrl":"10.1016/j.knosys.2025.113180","url":null,"abstract":"<div><div>Recent advances in Human-Computer Negotiation (HCN) systems have shown promising progress in simulating complex negotiation dialogues. However, these systems still need to grapple with critical challenges such as limited adaptability to proper negotiation tones, ineffective capture of long-term dependencies, and constraints in generating diverse, natural, and strategically sound responses. To address these limitations, we propose the HCN-RLR-CAN, a novel HCN model based on Round-Level Recurrence (RLR) and Causal Attention Network (CAN). Our approach, which uniquely processes dialogues as role-based parallel texts and employs a hierarchical encoder to comprehend user dialogue history, offers a fresh perspective on addressing these challenges. The model employs causal attention learning modules to separately model linguistic strategies and dialogue acts. Additionally, it employs a round-level recursive decoding mechanism that generates responses by synthesising historical dialogue information, dialogue act and strategy encodings, and previous decoding states. With its significant performance advantages over baseline models, the HCN-RLR-CAN model has the potential to inspire a new wave of research and development in the field of HCN systems.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"314 ","pages":"Article 113180"},"PeriodicalIF":7.2,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143480693","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

Topic modeling through rank-based aggregation and LLMs: An approach for AI and human-generated scientific texts

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

Knowledge-Based Systems

Pub Date : 2025-02-20 DOI: 10.1016/j.knosys.2025.113219

Tuğba Çelikten , Aytuğ Onan

The increasing presence of AI-generated and human-paraphrased content in scientific literature presents new challenges for topic modeling, particularly in maintaining semantic coherence and interpretability across diverse text sources. Traditional topic modeling methods, such as Latent Dirichlet Allocation (LDA) and Non-Negative Matrix Factorization (NMF), often suffer from inconsistencies and diminished coherence when applied to heterogeneous sources. Recently, large language models (LLMs) have demonstrated potential for enhanced topic extraction, yet they frequently lack the stability and interpretability required for reliable deployment. In response to these limitations, we propose a novel, robust ensemble framework that integrates rank-based aggregation and LLM-powered topic extraction to achieve consistent, high-quality topic modeling across AI-generated, AI-paraphrased, and human-generated scientific abstracts.

Our framework employs a rank-based aggregation scheme to reduce inconsistencies in LLM outputs and incorporates neural topic models to enhance coherence and semantic depth. By combining the strengths of traditional models and LLMs, our framework consistently outperforms baseline methods in terms of topic coherence, diversity, and stability. Experimental results on a diverse dataset of scientific abstracts demonstrate a substantial improvement in coherence scores and topic interpretability, with our ensemble approach outperforming conventional models and leading neural topic models by significant margins. This framework not only addresses the challenges of cross-source topic modeling but also establishes a benchmark for robust, scalable analysis of scientific literature spanning AI and human narratives.

{"title":"Topic modeling through rank-based aggregation and LLMs: An approach for AI and human-generated scientific texts","authors":"Tuğba Çelikten , Aytuğ Onan","doi":"10.1016/j.knosys.2025.113219","DOIUrl":"10.1016/j.knosys.2025.113219","url":null,"abstract":"<div><div>The increasing presence of AI-generated and human-paraphrased content in scientific literature presents new challenges for topic modeling, particularly in maintaining semantic coherence and interpretability across diverse text sources. Traditional topic modeling methods, such as Latent Dirichlet Allocation (LDA) and Non-Negative Matrix Factorization (NMF), often suffer from inconsistencies and diminished coherence when applied to heterogeneous sources. Recently, large language models (LLMs) have demonstrated potential for enhanced topic extraction, yet they frequently lack the stability and interpretability required for reliable deployment. In response to these limitations, we propose a novel, robust ensemble framework that integrates rank-based aggregation and LLM-powered topic extraction to achieve consistent, high-quality topic modeling across AI-generated, AI-paraphrased, and human-generated scientific abstracts.</div><div>Our framework employs a rank-based aggregation scheme to reduce inconsistencies in LLM outputs and incorporates neural topic models to enhance coherence and semantic depth. By combining the strengths of traditional models and LLMs, our framework consistently outperforms baseline methods in terms of topic coherence, diversity, and stability. Experimental results on a diverse dataset of scientific abstracts demonstrate a substantial improvement in coherence scores and topic interpretability, with our ensemble approach outperforming conventional models and leading neural topic models by significant margins. This framework not only addresses the challenges of cross-source topic modeling but also establishes a benchmark for robust, scalable analysis of scientific literature spanning AI and human narratives.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"314 ","pages":"Article 113219"},"PeriodicalIF":7.2,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143511340","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

Using machine learning in combinatorial optimization: Extraction of graph features for travelling salesman problem

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

Knowledge-Based Systems

Pub Date : 2025-02-20 DOI: 10.1016/j.knosys.2025.113216

Petr Stodola, Radomír Ščurek

Machine learning has emerged as a paradigmatic approach for addressing complex problems across various scientific disciplines, including combinatorial optimization. This article specifically explores the application of machine learning to the Travelling Salesman Problem (TSP) as a technique for evaluating and classifying graph edges. The methodology involves extracting a set of graph features and statistical measures for each edge in the graph. Subsequently, a machine learning model is constructed using the training data, and this model is employed to classify edges in a TSP instance, determining whether they are part of the optimal solution or not. This article contributes to existing knowledge in these key aspects: (a) enhancement of statistical measures, (b) introduction of a novel graph feature, and (c) preparation of training data to simulate real-world problem scenarios. Rigorous experimentation on benchmark instances from the well-established TSP library demonstrates a noteworthy increase in classification accuracy compared to the original approach without the improvements; various popular machine learning techniques are employed and evaluated. Furthermore, the characteristics and effects of the novel approaches are assessed and discussed, including their application to a basic heuristic algorithm. This research finds practical applications in problem reduction, involving the elimination of decision variables, or as a support for heuristic or metaheuristic algorithms in finding solutions.

{"title":"Using machine learning in combinatorial optimization: Extraction of graph features for travelling salesman problem","authors":"Petr Stodola, Radomír Ščurek","doi":"10.1016/j.knosys.2025.113216","DOIUrl":"10.1016/j.knosys.2025.113216","url":null,"abstract":"<div><div>Machine learning has emerged as a paradigmatic approach for addressing complex problems across various scientific disciplines, including combinatorial optimization. This article specifically explores the application of machine learning to the Travelling Salesman Problem (TSP) as a technique for evaluating and classifying graph edges. The methodology involves extracting a set of graph features and statistical measures for each edge in the graph. Subsequently, a machine learning model is constructed using the training data, and this model is employed to classify edges in a TSP instance, determining whether they are part of the optimal solution or not. This article contributes to existing knowledge in these key aspects: (a) enhancement of statistical measures, (b) introduction of a novel graph feature, and (c) preparation of training data to simulate real-world problem scenarios. Rigorous experimentation on benchmark instances from the well-established TSP library demonstrates a noteworthy increase in classification accuracy compared to the original approach without the improvements; various popular machine learning techniques are employed and evaluated. Furthermore, the characteristics and effects of the novel approaches are assessed and discussed, including their application to a basic heuristic algorithm. This research finds practical applications in problem reduction, involving the elimination of decision variables, or as a support for heuristic or metaheuristic algorithms in finding solutions.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"314 ","pages":"Article 113216"},"PeriodicalIF":7.2,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143487821","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

Multi-agent collaborative operation planning via cross-domain transfer learning

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

Knowledge-Based Systems

Pub Date : 2025-02-19 DOI: 10.1016/j.knosys.2025.113172

Cheng Ding , Zhi Zheng

Transfer learning has shown promising potentials in assisting multi-agent systems (MAS) to deal with complex collaborative tasks. In this work, we investigate MAS collaboration in 3D underwater environment. In response to the problem of high sampling cost in underwater operation when multi-agent without any prior knowledge, the multi-agent collaborative operation planning via cross-domain transfer learning (CDTL) is proposed. In CDTL, the training process of MAS is accelerated through learning the domain invariant knowledge from the samples of 2D ground collaborative tasks that easily obtained. First, the samples in ground tasks are divided into six state phases based on the semantic order of task execution, and a state transition graph is constructed accordingly. Then, a domain adaptation method with inter-class relationship (ICDA) is proposed, which focuses on the invariant semantic structure of the ground (source) and the underwater (target) task to capture prior knowledge. During the knowledge transferring, ICDA is used to correct decision of the agents’ policies that based on MAX-Q controller. Finally, the extensive experiments show that CDTL reduces the cost of physical time by 37.3% when the MAS completes the new task for the first time.

{"title":"Multi-agent collaborative operation planning via cross-domain transfer learning","authors":"Cheng Ding , Zhi Zheng","doi":"10.1016/j.knosys.2025.113172","DOIUrl":"10.1016/j.knosys.2025.113172","url":null,"abstract":"<div><div>Transfer learning has shown promising potentials in assisting multi-agent systems (MAS) to deal with complex collaborative tasks. In this work, we investigate MAS collaboration in 3D underwater environment. In response to the problem of high sampling cost in underwater operation when multi-agent without any prior knowledge, the multi-agent collaborative operation planning via cross-domain transfer learning (CDTL) is proposed. In CDTL, the training process of MAS is accelerated through learning the domain invariant knowledge from the samples of 2D ground collaborative tasks that easily obtained. First, the samples in ground tasks are divided into six state phases based on the semantic order of task execution, and a state transition graph is constructed accordingly. Then, a domain adaptation method with inter-class relationship (ICDA) is proposed, which focuses on the invariant semantic structure of the ground (source) and the underwater (target) task to capture prior knowledge. During the knowledge transferring, ICDA is used to correct decision of the agents’ policies that based on MAX-Q controller. Finally, the extensive experiments show that CDTL reduces the cost of physical time by 37.3% when the MAS completes the new task for the first time.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"314 ","pages":"Article 113172"},"PeriodicalIF":7.2,"publicationDate":"2025-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143474491","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

首页上一页

下一页尾页

类型

全部化学•材料生命科学医学物理工程技术环境•农林材料科学地球科学法学管理学化学环境科学与生态学计算机科学教育学经济学农林科学人文科学生物学数学物理与天体物理心理学综合性期刊其他工业工程理学历史学农学文学信息工程

数据库

全部 ACS Publications Elsevier ieeexplore Springer The Royal Society of Chemistry Wiley

期刊

Knowledge-Based Systems

全部 Acc. Chem. Res. ACS Applied Bio Materials ACS Appl. Electron. Mater. ACS Appl. Energy Mater. ACS Appl. Mater. Interfaces ACS Appl. Nano Mater. ACS Appl. Polym. Mater. ACS BIOMATER-SCI ENG ACS Catal. ACS Cent. Sci. ACS Chem. Biol. ACS Chemical Health & Safety ACS Chem. Neurosci. ACS Comb. Sci. ACS Earth Space Chem. ACS Energy Lett. ACS Infect. Dis. ACS Macro Lett. ACS Mater. Lett. ACS Med. Chem. Lett. ACS Nano ACS Omega ACS Photonics ACS Sens. ACS Sustainable Chem. Eng. ACS Synth. Biol. Anal. Chem. BIOCHEMISTRY-US Bioconjugate Chem. BIOMACROMOLECULES Chem. Res. Toxicol. Chem. Rev. Chem. Mater. CRYST GROWTH DES ENERG FUEL Environ. Sci. Technol. Environ. Sci. Technol. Lett. Eur. J. Inorg. Chem. IND ENG CHEM RES Inorg. Chem. J. Agric. Food. Chem. J. Chem. Eng. Data J. Chem. Educ. J. Chem. Inf. Model. J. Chem. Theory Comput. J. Med. Chem. J. Nat. Prod. J PROTEOME RES J. Am. Chem. Soc. LANGMUIR MACROMOLECULES Mol. Pharmaceutics Nano Lett. Org. Lett. ORG PROCESS RES DEV ORGANOMETALLICS J. Org. Chem. J. Phys. Chem. J. Phys. Chem. A J. Phys. Chem. B J. Phys. Chem. C J. Phys. Chem. Lett. Analyst Anal. Methods Biomater. Sci. Catal. Sci. Technol. Chem. Commun. Chem. Soc. Rev. CHEM EDUC RES PRACT CRYSTENGCOMM Dalton Trans. Energy Environ. Sci. ENVIRON SCI-NANO ENVIRON SCI-PROC IMP ENVIRON SCI-WAT RES Faraday Discuss. Food Funct. Green Chem. Inorg. Chem. Front. Integr. Biol. J. Anal. At. Spectrom. J. Mater. Chem. A J. Mater. Chem. B J. Mater. Chem. C Lab Chip Mater. Chem. Front. Mater. Horiz. MEDCHEMCOMM Metallomics Mol. Biosyst. Mol. Syst. Des. Eng. Nanoscale Nanoscale Horiz. Nat. Prod. Rep. New J. Chem. Org. Biomol. Chem. Org. Chem. Front. PHOTOCH PHOTOBIO SCI PCCP Polym. Chem.

﹀