Weighted Clustering of Time Series

Abstract

We will describe here a method for the clustering of time series. This method does not give the same importance to all of the observations; instead, it lets the most important observations, for instance the most recent, have a larger weight. A fundamental problem in the clustering of time series is the choice of a relevant metric, and here, we will use a metric, based on Pearson’s correlation coefficient, which uses the notion of weighted mean and weighted covariance. We present also some motivating applications.