Formalism for a multiresolution time series database model

Abstract

We formalise a specialised database management system model for time series using a multiresolution approach. These special purpose database systems store time series lossy compressed in a space-bounded storage. Time series can be stored at multiple resolutions, using distinct attribute aggregations and keeping its temporal attribute managed in a consistent way.

The model exhibits a generic approach that facilitates its customisation to suit better the actual application requirements in a given context. The elements, the meaning of which depends on a real application, are of generic nature.

Furthermore, we consider some specific time series properties that are a challenge in the multiresolution approach. We also describe a reference implementation of the model and introduce a use case based on real data.