Period analysis based on SVM and wavelet variance for time series

Autor: Qibin Fan, Ruhua Cai
Rok vydání: 2010
Předmět:
Zdroj: 2010 International Conference on Intelligent Computing and Integrated Systems.
DOI: 10.1109/iciss.2010.5656104
Popis: In the time series period analysis, the period obtained by the maximal wavelet variance has serious errors. In this paper, we present a method of LS-SVM to approximate the wavelet variance or power spectrum at different scales, and then obtain the period of sequence by estimating the maximum value. The experiment indicates that LS-SVM method can approximate wavelet variance effectively, and can estimate the period of the time series accurately. It is an effective method for time series period analysis and power spectral analysis.
Databáze: OpenAIRE