Paper 4

Efficient Mining of Lag Patterns in Evolving Time Series

Authors: Dhaval Patel, Wynne Hsu, and Mong Li Lee

Volume 11 (2013)

Abstract

Time series motifs are sets of similar subsequences. Lag patterns, or the invariant ordering among time series motifs, depict localized repeated associative relationships across multiple real valued time series. Lag patterns are of special interest inmany real world applications, such as constructing stock portfolio in financial domain, extracting regulator-target relationship in bioinformatics domain, etc. However, mining lag patterns is computationally intensive, particularly in evolving time series data. In this paper, we present an efficient algorithm called LPMiner? that iteratively discovers motifs and generates lag patterns of increasing length. We also design an incremental algorithm called incLPMiner to mine lag patterns in the presence of frequent database updates. Experimental analysis on real world time series datasets demonstrate the efficiency and scalability of our proposed algorithms.