An Analysis On Asynchronous Periodic Pattern Detection In Multi Variate Time Series

Abstract

Data mining is an area that emphases on mining patterns from accumulated data in order to gain hidden and useful knowledge. Sequential pattern extraction has become the thrust area as it has better capability to predict the future based on past experiences. Periodic patterns are type of patterns that can be mined from sequences as they repeat regularly with a specific period in a given sequence. Patterns whose successive repetitions are perfectly aligned with a periodicity are called as synchronous periodic patterns. The multivariate time series is transformed into symbol sequence(s) by representing the variable-value pairs observed at each timestamp by the symbols of the best “n” representative closed interesting subspaces applicable at that timestamp