Similarity measure for time series based on piecewise linear approximation

Time series data have attracted much research recently. Among many valuable research topics, similarity measure is very important. The choice of similarity measure can affect the result of data mining tasks. Piecewise linear approximation is a kind of dimensionality reduction technique, based on piecewise linear approximation, we introduce two similarity measures for time series, which satisfy four properties of metric. We use the measures in similarity search problem, experiments on real data show their effectiveness.

I. Introduction

A time series is a sequence of observations, which are often measured at successive time. Time series data emerge in various domains, including financial stock analysis, medical diagnosis, sensor network monitoring, weather forecasting. The pervasiveness and importance of time series data has made them attract more and more researchers' attention. Time series data have even been considered as one of ten challenging problems in data mining research [1].