Developing Fast Techniques for Periodicity Analysis of Time Series

Abstract

The periodicity analysis of time series is the analysis of a set of measurements recorded for one or more variables arranged over time. Periodicity mining can be used in many application domains such as meteorology, astronomy, and econometrics. The techniques developed to find the periodicity are based on Fourier transform (FT), wavelet transform (WT), and dynamic time warping (DTW). Although the DTW performs well for periodicity analysis of time series, it is computationally complex. This study proposes two new algorithms which are formed by combining fast Fourier transform (FFT) with dynamic time warping (FFT-DTW) and wavelet transform with dynamic time warping (WT-DTW). The performances of the proposed algorithms were evaluated on real and synthetic datasets and the results are promising.