Change Detection of Model Transitions in Proxel Based Simulation of CHnMMs
To analyze discrete stochastic models, Virtual Stochastic Sensors were developed at the Otto-von-Guericke-University Magdeburg. This procedure makes it possible to reconstruct the behavior of a broader class of hidden models, like Conversive Hidden non-Markovian Models, in a very efficient way. One assumption of this approach is that the distribution functions, which describe the state changes of the system, are time-homogeneous. However, this assumption is not always true when it comes to real world problems.
To overcome this limitation, the paper presents an algorithm where the concept of Virtual Stochastic Sensors was extended with statistical tests to continuously evaluate the parameters of a Conversive Hidden non-Markovian Model and the current results. If needed, the tests stop the execution of the behavior reconstruction and reevaluate the model based on the current knowledge about the system.
The project showed that detecting the change and adjusting the model is possible during the behavior reconstruction, improving reconstruction accuracy. The method was tested using four types of distribution functions, three of which showed very good results. By using this new algorithm, one is able to construct adaptive models for behavior reconstruction without additional conceptual effort. In this way, loss of modeling accuracy due to abstractions in the modeling process can be balanced. Another possible application appears in the case of long time investigations. The change detection method can be invoked after a given period to reevaluate the system model and make the relevant adjustments if needed.
KeywordsVirtual Stochastic Sensor Conversive Hidden non-Markovian Model Drift Change detection
- 3.Buchholz, R.: Conversive Hidden non-Markovian Models. Otto-von-Guericke-Universität, Magdeburg (2012)Google Scholar
- 4.Buchholz, R., Krull, C., Horton, G.: Virtual stochastic sensors: how to gain insight into partially observable discrete stochastic systems. In: The 30th IASTED International Conference on Modelling (2011)Google Scholar
- 7.Horton, G.: A new paradigm for the numerical simulation of stochastic petri nets with general firing times. In: Proceedings of the European Simulation Symposium, ESS 2002 (2002)Google Scholar
- 8.Krull, C., Horton, G.: Hidden non-Markovian Models: formalization and solution approaches. In: 6th Vienna International Conference on Mathematical Modelling (2009)Google Scholar
- 9.Krull, C., Horton, G., Denkena, B., Dengler, B.: Virtual stochastic sensors for reconstructing job shop production workflows. In: 2013 8th EUROSIM Congress on Modelling and Simulation (EUROSIM), pp. 276–281 (2013)Google Scholar
- 10.Lazarova-Molnar, S.: The Proxel-Based Method: Formalisation, Analysis and Applications. Otto-von-Guericke-Universität, Magdeburg (2005)Google Scholar
- 12.NIST: NIST/SEMATECH e-Handbook of Statistical Methods (2013). http://www.itl.nist.gov/div898/handbook/index.htm
- 13.Wilson, E.: Virtual sensor technology for process optimization (1997). Presentation at the ISSCAC 1997Google Scholar