Please use this identifier to cite or link to this item:
Title: Comparative Analysis of Noisy Time Series Clustering
Authors: Кіріченко, Л. О.
Радівілова, Т. А.
Ткаченко, А. Є.
Keywords: Time Series Clustering
Atypical Time Series
Noisy Time Series Clustering
Issue Date: 2019
Citation: Kirichenko L. Comparative Analysis of Noisy Time Series Clustering / L. Kirichenko L., T. Radivilova, A. Tkachenko // Computational Linguistics and Intelligent Systems : proceedings of the 3rd International Conference, April 18-19, 2019. – Kharkiv, 2019. – Volume I. – P.84-196.
Abstract: A comparative analysis of the clustering of sample time series was performed. The clustering sample contained time series of various types, among which atypical objects were present. In the numerical experiment, white noise with different variance was added to the time series. Clustering was performed by k-means and DBSCAN methods using various similarity functions of time series. The values of the quality functionals were quantitative measures of the quality of clustering. The best results were shown by the DBSCAN method using the Euclidean metric with a Complexity Invariant Distance. The method allows to separate a cluster with atypical series at different levels of additive noise. The results of the clustering of real time series confirmed the applicability of the DBSCAN method for detecting anomaly.
Appears in Collections:Кафедра прикладної математики (ПМ)

Files in This Item:
File Description SizeFormat 
Kirich_2019.pdf452.27 kBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.