Wavelet Transform and Statistical Characteristics Extraction Applied to Times Series Clustering Analysis
Liu Hanl · 2014
The functionality of the multi-resolution of wavelet transform is utilized and a multiple wavelet transform is performed on time series data in this paper. By wavelet transform, the time series can be decomposed into scale components and detail components. By retaining the scale components and removing the detail components, the trend information of time series is extracted. And then, the statistical characteristics data are calculated from the time series. In order to perform clustering analysis on the time series, the trend data and the statistical characteristics of time series are combined to form the input vector of SOM neural network. After that, a series of experiments for clustering analysis is conducted on four sets of simulated time series and one set of MODIS remote sensing vegetation data. The results of clustering analysis show that the proposed method is effective.