To search, Click below search items.


All Published Papers Search Service


Noise Resilient Periodicity Mining in Time Series Data Bases


K.Prasad Rao, M.Gayathri


Vol. 14  No. 7  pp. 41-44


One important aspect of the data mining is found out the interesting periodic patterns from time series databases. One area of research in time series databases is periodicity detection. Time series prognosticating is the use of a model to presage future values based on previously observed values. Discovering periodic patterns from an entire time series has proven to be inadequate in applications where the periodic patterns occur only with in small segments of the time series. In this paper we proposed a couple of algorithms to extract interesting periodicities (Symbol, Sequence, and Segment) of whole time series or part of the time series. One algorithm utilizes Dynamic warping Technique and other one is suffix tree as a data structure. Couple of these algorithms has been successfully incontestable to work with replacement, Insertion, deletion, or a mixture of these types of noise. The algorithm is compared with DTW (Dynamic Time Warping) algorithm on different types of data size and periodicity on both real and synthetic data. The worst case complexity of the suffix tree noise resilient algorithm is O (n.k2) where n is the maximum length of the periodic pattern


Time Series, suffix tree, periodicity detection, noise resilient, DTW