An Architectural Model for High Performance Pattern Matching in Linked Historical Data

In times of global digitalization and interconnectedness the virtual Cyber Physical Systems (CPS) are getting more and more on importance. These CPS and their relations among themselves can be investigated using appropriate data acquired by the inherent sensors. The multivariate, multiscale, multimodal sensor data can be modeled and analyzed as a dynamically evolving spatio-temporal complex network. These graphs as well as the patterns estimated in historical data can then be used for real time comparison with momentary computed patterns. Therefore providing linked data from memory is an important need to accomplish real time constraints especially in case of CPS in critical medical systems. Since the handling of graphs in the traditional relational database systems is problematic an encouraging approach is the storage of these data in graph databases which are appropriate for the handling of linked data. Therefore we propose the graph database Neo4J and demonstrate first applications of the approach within medical use-cases.