A methodological trajectory has been described dealing with the `Novelty' or `Surprise' issue in time series records arising from real world complex systems. It is based on extracting regularity (or scaling) characteristics of non-differentiable time series with the wavelet transform, on modelling the complex system using multi-fractal properties and on investigating novelty in the context of the possible non-stationarity of such a model.

Springer
Advances in soft computing
Intelligent Information Systems - New Trends in Intelligent Processing and Web Mining
Database Architectures

Struzik, Z. R. (2003). Taming Surprises. In Intelligent Processing and Web Mining 2003 (4). Springer.