Article
Computer Science, Artificial Intelligence
Dalibor Krleza, Boris Vrdoljak, Mario Brcic
Summary: Anomaly detection in data streams is typically solved in the online phase, while good macro-clustering is produced in the offline phase, making it challenging for two-phase clustering algorithms to equally excel in both anomaly detection and macro-clustering. The proposed statistical hierarchical clustering algorithm aims to address this issue, by using statistical inference on input data streams and constantly updating statistical distributions to adaptively classify data without the need for prior information on the number of clusters or outliers. Testing against typical clustering algorithms demonstrated the universality and quality of the proposed algorithm.
Article
Chemistry, Multidisciplinary
Young Jong Song, Ki Hyun Nam, Il Dong Yun
Summary: Surface-mounted device (SMD) assembly machines are production lines that assemble various products for specific purposes. With the increasing diversity of required products, the models for product anomaly detection are also growing linearly. This paper demonstrates the use of latent vectors obtained from an autoencoder model to handle a large number of new products. By hierarchically clustering the latent vectors, the model can identify product groups with similar characteristics for efficient supervision. The experimental results show that this anomaly detection method using hierarchical clustering of latent vectors is practical for SMD anomaly detection management.
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Artificial Intelligence
Riccardo Patriarca, Francesco Simone, Giulio Di Gravio
Summary: This article proposes a novel method based on Machine Learning to detect forecasting anomalies in historic data and use them to predict potential threats in future aerodrome forecasts. The method aims to enhance decision makers' ability to manage aerodrome weather forecasting and understand critical factors related to their accuracy levels.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Transportation Science & Technology
Weizun Zhao, Lishuai Li, Sameer Alam, Yanjun Wang
Summary: This study proposes an incremental anomaly detection method based on Gaussian Mixture Model for identifying common patterns and detecting outliers in flight operations from digital flight data. Compared to traditional offline GMM methods, this approach showed significant improvements in processing time and memory usage, indicating that the incremental learning scheme is effective in dealing with dynamically growing data in flight data analytics.
TRANSPORTATION RESEARCH PART C-EMERGING TECHNOLOGIES
(2021)
Article
Computer Science, Information Systems
Md. Mumtahin Habib Ullah Mazumder, Md. Eusha Kadir, Sadia Sharmin, Md. Shariful Islam, Muhammad Mahbub Alam
Summary: The recent trend in network intrusion detection is to use key features of machine learning algorithms to detect network traffic anomalies. Network traffic contains high dimensional features that significantly affect data-driven approaches. Therefore, the performance of machine learning-based approaches depends on the appropriate set of network data features. Different feature selection and extraction methods are used to obtain informative and compact feature sets. However, existing methods often fail to achieve expected performance due to the inability to effectively remove redundant features and incorporate features with complementary information. In this paper, we propose a cluster-based feature extraction method using Mahalanobis distance (cFEM) that clusters correlated features and extracts new feature representations based on a distance metric. The extracted features are employed to train different machine learning classifiers. Extensive experiments on renowned datasets demonstrate that cFEM outperforms state-of-the-art intrusion detection methods in various performance metrics such as detection rate (99.61%) and false alarm rate (0.26%). Further experiments show that the extracted features are discriminative, non-redundant, and able to capture complementary information.
INTERNATIONAL JOURNAL OF INFORMATION SECURITY
(2023)
Article
Engineering, Mechanical
Lin-Feng Mei, Wang-Ji Yan, Ka-Veng Yuen, Wei-Xin Ren, Michael Beer
Summary: This paper proposes a new damage detection method by integrating the advantage of transmissibility function (TF) as a health index sensitive to damage but robust to excitation and agglomerative hierarchical clustering (AHC) with intuitive explanation and visualization but avoiding specifying the number of clusters. The method utilizes a multivariate probabilistic distance-based similarity metric to account for the uncertainty and correlation of multiple TFs. The performance of the method has been validated through case studies and it shows better performance compared to traditional methods.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Computer Science, Information Systems
Beiji Zou, Kangkang Yang, Xiaoyan Kui, Jun Liu, Shenghui Liao, Wei Zhao
Summary: This paper proposes an unsupervised anomaly detection algorithm (GC-ADS) based on grid clustering and Gaussian distribution, which can accurately and quickly detect anomalies in real-time and evolving streaming data. The data space is segmented using a grid structure, and data points are mapped to grids and clustered to preliminarily judge anomalies based on cluster density. A noise recognition model based on data similarity and Gaussian distribution, as well as a data filtering model based on grid and sliding window, are designed to save memory and retain valid information. Experimental results show that GC-ADS detects anomalies more accurately with lower time cost compared to other methods on the Numenta anomaly benchmark.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
J. Miguel Salazar, Pablo Lopez-Ramirez, Oscar S. Siordia
Summary: This article presents an approach for detecting crowd activity in geographic point data by combining hierarchical scale structures with density-based clustering algorithms. The method includes generating synthetic data and improving automatic parameter selection algorithms.
PEERJ COMPUTER SCIENCE
(2022)
Article
Computer Science, Information Systems
J. Fumanal-Idocin, I Rodriguez-Martinez, A. Indurain, M. Minarova, H. Bustince
Summary: This paper proposes a simulation-based anomaly detection algorithm that identifies abnormal observations significantly different from normal ones using the aggregation of gravitational forces and cluster analysis, without prior knowledge or data labels.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Information Systems
En-Hau Yeh, Phone Lin, Ming-Wey Huang
Summary: This article proposes an anomaly detection framework based on population distribution, using mobile network log data to monitor real-time population mobility patterns and identify critical indicators for sudden events. The framework shows a high practicality in actual situations, as demonstrated by the experiments conducted during the 2018 Hualien Earthquake in Taiwan.
IEEE INTERNET OF THINGS JOURNAL
(2022)
Article
Automation & Control Systems
Jingyu Yang, Zuogong Yue
Summary: Multivariate time series anomaly detection is crucial but challenging in complex industrial processes. This article proposes a novel method called HiSTAR, which overcomes the limitations of existing approaches and achieves superior anomaly detection performance by learning hierarchical normality-enclosing hyperspheres. HiSTAR also provides consistent anomaly localization results. Experimental results from three industrial case studies validate the effectiveness of the proposed method.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Amna Habib, Muhammad Akram, Cengiz Kahraman
Summary: This paper proposes a graph theory-based agglomerative hierarchical clustering technique for Pythagorean fuzzy sets. By considering uncertain parameters and qualitative aspects in the expression, the algorithm demonstrates practicality and efficiency in clustering problems in complex networks.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Nana Liu, Zeshui Xu, Xiao-Jun Zeng, Peijia Ren
Summary: This paper introduces a new method for clustering LOR information using the AHC algorithm, by extending existing distance measure methods and simplifying aggregation methods. A numerical case study is presented to illustrate the algorithm's usage, and discussions are made on the features of the algorithm.
INFORMATION SCIENCES
(2021)
Article
Chemistry, Multidisciplinary
Jui-Hung Liu, Nelson T. Corbita, Rong-Mao Lee, Chun-Chieh Wang
Summary: This study uses the Mahalanobis distance to detect anomalies in wind turbine operation and finds that it can detect anomalies in different components. This analysis can be used in condition monitoring systems of wind turbines.
APPLIED SCIENCES-BASEL
(2022)
Article
Chemistry, Analytical
Claudia Malzer, Marcus Baum
Summary: This article explores the applicability of HDBSCAN for radar measurements clustering, proposing cluster-level constraints based on cluster candidates and a distance threshold to improve results. Experiments on datasets such as nuScenes show that cluster-level constraints can significantly enhance the performance of HDBSCAN compared to the unsupervised method, but careful selection of constraint criteria is necessary.