Pattern Recognition of Time Series using Wavelets
Elizabeth Ann Maharaj · COMPSTAT · 2002
In this paper a pattern recognition procedure for time series using wavelets is developed. This is done by means of a randomization test based on the ratio of the sum of squared wavelet coefficients of pairs of time series at different scales. A simulation study using pairs of stationary and non-stationary time series and using the Haar and Daubechies wavelets reveals that the test performs fairly well at scales where there are a sufficient number of wavelet coefficients. The test is applied to a set of financial time series.