Scholarlink

Abstract <p>This work provides a comparative analysis of the anomaly detection methods in time series based on machine learning and spectral analysis. Four approaches are considered: the <i>K</i>-nearest neighbors, the singular spectrum analysis, and prediction by means of two models, exponential and linear regression. The methods are tested on synthetic time series imitating data with an addition of artificial anomalies. The efficiency of the algorithms is evaluated by the Precision, Recall, and F1-score metrics. The results of research demonstrate the advantages and limitations of each method in different scenarios of time series analysis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparison of Anomaly Detection Algorithms Based on Machine Learning and Spectral Analysis Methods

  • 期刊论文

  • 发表日期：

    2025年9月1日

查看全文
  • D. A. Makeev,
  • K. E. Karavaeva,
  • P. A. Fadeeva,
  • N. E. Shapkina,
  • P. V. Golubtsov,
  • A. I. Chulichkov

摘要

Abstract

This work provides a comparative analysis of the anomaly detection methods in time series based on machine learning and spectral analysis. Four approaches are considered: the K-nearest neighbors, the singular spectrum analysis, and prediction by means of two models, exponential and linear regression. The methods are tested on synthetic time series imitating data with an addition of artificial anomalies. The efficiency of the algorithms is evaluated by the Precision, Recall, and F1-score metrics. The results of research demonstrate the advantages and limitations of each method in different scenarios of time series analysis.