Using suffix trees for periodicity detection in time series databases

Periodicity detection in time series has been used extensively for predicting trends in time series databases, such as weather data, stock market, etc. In this paper, we approach periodicity detection using the suffix tree as the underlying data structure. Our algorithm not only discovers the periodicity of a single symbol or of the entire series, called segment periodicity, but can also detect the sequence (multiple-symbol) periodicity. Unlike others, our algorithm uses various period pruning approaches that result in producing more meaningful non-redundant periods. The developed methodology has been validated by conducting a number of experiments for testing its applicability and effectiveness compared to the other similar approaches.