A proposal for the hierarchical segmentation of time series. Application to trend-based linguistic description

In this paper we propose methods for obtaining hierarchical segmentations of time series on the basis of the Iterative End-Point Fit Algorithm. We discuss on the utility of the methods for different cases. We illustrate the usefulness of the hierarchical segmentations with an application in linguistic description of trends in time series. A linguistic description based on a segmentation of the time series that do not necessarily corresponds to a level of the hierarchy is obtained by describing segments in different levels that form a segmentation satisfying a quality model.