The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Dalam senario Kecerdasan Buatan untuk Operasi IT, KPI (Petunjuk Prestasi Utama) ialah penunjuk pemantauan operasi dan penyelenggaraan yang sangat penting, dan penyelidikan mengenai pengesanan anomali KPI juga telah menjadi tumpuan sejak beberapa tahun kebelakangan ini. Mensasarkan masalah kecekapan pengesanan rendah dan pembelajaran perwakilan yang tidak mencukupi bagi kaedah sedia ada, kertas ini mencadangkan kaedah pengesanan anomali KPI berasaskan kluster pantas HCE-DWL. Kertas kerja ini mula-mula mengguna pakai gabungan pengelompokan aglomeratif hierarki (HAC) dan penugasan mendalam berdasarkan CNN-Embedding (CE) untuk melaksanakan analisis kelompok (iaitu HCE) pada data KPI, untuk meningkatkan kecekapan pengelompokan data KPI, dan kemudian secara berasingan pusat setiap kelompok KPI dan Transformed Outlier Scores (TOS) diberi pemberat, dan akhirnya ia dimasukkan ke dalam model LightGBM untuk pengesanan (model Double Weight LightGBM, dirujuk sebagai DWL). Melalui analisis eksperimen perbandingan, terbukti bahawa algoritma boleh meningkatkan kecekapan dan ketepatan pengesanan anomali KPI dengan berkesan.
Yun WU
Northeast Electric Power University
Yu SHI
Northeast Electric Power University
Jieming YANG
Northeast Electric Power University
Lishan BAO
Industry Group Co. Ltd.
Chunzhe LI
State Grid Jilin Electric Power Co. Ltd.
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Salinan
Yun WU, Yu SHI, Jieming YANG, Lishan BAO, Chunzhe LI, "A KPI Anomaly Detection Method Based on Fast Clustering" in IEICE TRANSACTIONS on Communications,
vol. E105-B, no. 11, pp. 1309-1317, November 2022, doi: 10.1587/transcom.2021TMP0002.
Abstract: In the Artificial Intelligence for IT Operations scenarios, KPI (Key Performance Indicator) is a very important operation and maintenance monitoring indicator, and research on KPI anomaly detection has also become a hot spot in recent years. Aiming at the problems of low detection efficiency and insufficient representation learning of existing methods, this paper proposes a fast clustering-based KPI anomaly detection method HCE-DWL. This paper firstly adopts the combination of hierarchical agglomerative clustering (HAC) and deep assignment based on CNN-Embedding (CE) to perform cluster analysis (that is HCE) on KPI data, so as to improve the clustering efficiency of KPI data, and then separately the centroid of each KPI cluster and its Transformed Outlier Scores (TOS) are given weights, and finally they are put into the LightGBM model for detection (the Double Weight LightGBM model, referred to as DWL). Through comparative experimental analysis, it is proved that the algorithm can effectively improve the efficiency and accuracy of KPI anomaly detection.
URL: https://global.ieice.org/en_transactions/communications/10.1587/transcom.2021TMP0002/_p
Salinan
@ARTICLE{e105-b_11_1309,
author={Yun WU, Yu SHI, Jieming YANG, Lishan BAO, Chunzhe LI, },
journal={IEICE TRANSACTIONS on Communications},
title={A KPI Anomaly Detection Method Based on Fast Clustering},
year={2022},
volume={E105-B},
number={11},
pages={1309-1317},
abstract={In the Artificial Intelligence for IT Operations scenarios, KPI (Key Performance Indicator) is a very important operation and maintenance monitoring indicator, and research on KPI anomaly detection has also become a hot spot in recent years. Aiming at the problems of low detection efficiency and insufficient representation learning of existing methods, this paper proposes a fast clustering-based KPI anomaly detection method HCE-DWL. This paper firstly adopts the combination of hierarchical agglomerative clustering (HAC) and deep assignment based on CNN-Embedding (CE) to perform cluster analysis (that is HCE) on KPI data, so as to improve the clustering efficiency of KPI data, and then separately the centroid of each KPI cluster and its Transformed Outlier Scores (TOS) are given weights, and finally they are put into the LightGBM model for detection (the Double Weight LightGBM model, referred to as DWL). Through comparative experimental analysis, it is proved that the algorithm can effectively improve the efficiency and accuracy of KPI anomaly detection.},
keywords={},
doi={10.1587/transcom.2021TMP0002},
ISSN={1745-1345},
month={November},}
Salinan
TY - JOUR
TI - A KPI Anomaly Detection Method Based on Fast Clustering
T2 - IEICE TRANSACTIONS on Communications
SP - 1309
EP - 1317
AU - Yun WU
AU - Yu SHI
AU - Jieming YANG
AU - Lishan BAO
AU - Chunzhe LI
PY - 2022
DO - 10.1587/transcom.2021TMP0002
JO - IEICE TRANSACTIONS on Communications
SN - 1745-1345
VL - E105-B
IS - 11
JA - IEICE TRANSACTIONS on Communications
Y1 - November 2022
AB - In the Artificial Intelligence for IT Operations scenarios, KPI (Key Performance Indicator) is a very important operation and maintenance monitoring indicator, and research on KPI anomaly detection has also become a hot spot in recent years. Aiming at the problems of low detection efficiency and insufficient representation learning of existing methods, this paper proposes a fast clustering-based KPI anomaly detection method HCE-DWL. This paper firstly adopts the combination of hierarchical agglomerative clustering (HAC) and deep assignment based on CNN-Embedding (CE) to perform cluster analysis (that is HCE) on KPI data, so as to improve the clustering efficiency of KPI data, and then separately the centroid of each KPI cluster and its Transformed Outlier Scores (TOS) are given weights, and finally they are put into the LightGBM model for detection (the Double Weight LightGBM model, referred to as DWL). Through comparative experimental analysis, it is proved that the algorithm can effectively improve the efficiency and accuracy of KPI anomaly detection.
ER -