Finding motifs of financial data streams in real time

Abstract

Finding motifs of financial data streams in real time is a very interesting and valuable work. We hope to find the motif existing in financial data streams on local trend subsequence. A stock market trader might use such a tool to spot arbitrage opportunities or escape the underlying venture. The paper introduces a novel distance measurement, that is SDD (Slope Duration Distance), for local subsequences. At the same time, we propose an efficient algorithm of motif discovery over a great deal of financial data streams, that is PMDGS (P-Motif Discovery based on Grid Structure), which make use of PLA (Piecewise Linear Approximation) technology and grid structure. Extensive experiments on synthetic data and real world financial trading data show that our model provides several orders of magnitude performance improvement relative to traditional naive linear scan techniques. © 2008 Springer Berlin Heidelberg.