Complex Data: Mining using Patterns

There is a growing need to analyse sets of complex data, i.e., data in which the individual data items are (semi-) structured collections of data themselves, such as sets of time-series. To perform such analysis, one has to redefine familiar notions such as similarity on such complex data types. One can do that either on the data items directly, or indi- rectly, based on features or patterns computed from the individual data items. In this paper, we argue that wavelet decomposition is a general tool for the latter approach