Simulation and modeling of semantically enriched time series

Abstract

Time series are a part of our day to day life. A lot of computation and data processing problems can be solved by time series processing. Therefore, the question arises how to model and simulate time series data, which contains semantics and meta-data on the one hand, and how to derive decisions out of this data on the other hand. A lot of data and data streams could be efficiently and reasonably represented as time series. This allows a simple and highly accurate reuse in real-time or time-shifted. It is useful and natural for a lot of different domains to store and manage data in this manner. The processing of time series is necessary to draw conclusions out of data and to implement user and application specific software solutions. This is why time series processing is an optimal solution for software in many different fields of application. Therefore, we have developed a generic language to process time series data, which supports homogeneous and heterogeneous time series, complex data structures, working with time patterns, time intervals and single slots, and complex calculation with predefined and user defined functions. The main advantages are high expressiveness, user-friendly syntax, good extensibility, and meaningful data models. Nowadays existing decision support systems and systems for time series processing still have some weaknesses. For example the fact that meta-information may still be missing or not integrated in the processing, that ontologies are not used, which means that contexts and connections could not be correctly recognized and respected, and particularly no option to bind domain-specific ontologies during run-time, which makes domain-specific processing hard to implement. A semantic-enabled time series processor could use predefined or user-generated ontologies to enrich information with the appropriate meaning. This would allow automatic consideration of domain-specific calculations and decision support, a low fault probability (as complex expressions are easier to be formulated), verification of meaningfulness and reasonability, and much more additional features. The models provided by this language could be integrated in interactive decision support systems for endusers. The advantage is that they are dynamic and there is no need of touching the DSS code or manually apply models to DSS. As the current field of application is Environmental Informatics, the dynamic is restricted to this field at the current state of implementation. The main advantage in this dynamic usage is the possibility to replace models as needed and to process multiple models at the same time (which saves computation time and resources). The fields of application are nearly endless. Whether in industrial measurement and control applications to provide a reliable processing of measurement data, in risk management applications for domain-crossing risk calculations, in environment monitoring applications for alerting and legally compliant reporting in different domains, in eHealth applications for linking of diagnosis-relevant data, such as subjective sense of well-being, or in meteorology. As this technology is part of the TaToo project, its primary field of application is TaToo and applications which use the TaToo Framework, as well as other projects, technologies and applications based on TaToo. However, semantic time series processing would be a promising supplement for every decision support system, and has the potential to contribute improvements to this scientific area.