Knowledge Discovery as Emergent Models
A complex system is a system of many interacting components or subsystems. Examples are: a living cell composed of molecules, an organism composed of cells, an ecosystem composed of organisms, a society composed of individuals, a large enterprise composed of employees and/or business units, and an economic system composed of enterprises. The behaviour of each component may be known, but the behaviour of the whole system can be difficult to understand, therefore unpredictable.
Using the Emergent Models methodology developed by Jacob Stolk, computer simulations of complex systems are combined with knowledge discovery algorithms to obtain macro-level models, called Emergent Models, of system behaviour emerging from behaviour and interactions of a system's components. Effectively, a micro-level simulation produces very detailed data on the behaviour and interactions of all fine-grained elements of the system. These data are analysed using a knowledge discovery algorithm, for example genetic programming. The extracted knowledge is encoded in a macro-level model of the whole system.
This methodology has considerable advantages. On a fundamental level, a macro-level model is obtained that is founded on an understanding of the micro-level behaviour of a system's components and their interactions rather than only empirical observations of the whole system. On an operational level, micro-level components of a complex system often have random elements in their behaviour, so every run of a micro-level simulation is different and any particular run is not guaranteed to produce results that are typical for the system. In contrast, an emergent model will represent the average behaviour of the system. Therefore, it enables more confident prediction of the system's future behaviour.
As a bonus, once the macro-level model is obtained, simulations of the system can be run orders of magnitude faster than simulations including all individual components of the system. Macro-level simulations producing the same system behaviour as micro-level simulations are typically run with a 1000 or more times better performance than a corresponding micro-level simulation.
The methodology is also applicable to the inverse problem of discovering micro-level models producing observed macro-level behaviour. For example, it has been used to discover possible genetic and biochemical networks explaining observed phenotype behaviour.
References
Stolk J. 2013. Constructing Water Tank Delivery Schedules through Combined Vehicle Routing and Packing Decisions. 22nd National Conference of the Australian Society for Operations Research and MODSIM2013 International Congress on Modelling and Simulation, Dec. 2013, Adelaide, Australia.
Stolk J. 2013. Combining Vehicle Routing and Packing for Optimal Delivery Schedules of Water Tanks. OR Insight 26(3): pp. 167-190.
Stolk J. 2009. Complex Systems Simulation for Risk Assessment in Flood Incident Management. In Anderssen R.S., R.D. Braddock and L.T.H. Newham (eds) 18th World IMACS Congress and MODSIM09 International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand and International Association for Mathematics and Computers in Simulation, July 2009, pp. 2377-2383. ISBN: 978-0-9758400-7-8.
Stolk J. and Hanan J. 2009. Emergent Models in a Multi-level Biochemical Network Regulating Pea Flowering. In Anderssen, R.S., R.D. Braddock and L.T.H. Newham (eds) 18th World IMACS Congress and MODSIM09 International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand and International Association for Mathematics and Computers in Simulation, July 2009, pp. 2377-2383. ISBN: 978-0-9758400-7-8.
Stolk H.J. and Hanan J. 2007. Discovering Genetic Regulatory Network Models in Pisum sativum. In Oxley L. and Kulasiri D. (eds) MODSIM 2007 International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand, December 2007, pp. 74-80. ISBN : 978-0-9758400-4-7.
Stolk H.J., Hanan J. and Zalucki M.P. 2007. Subpopulation Agents Emerge from Individual Agents in Metapopulation Simulations. In Oxley L. and Kulasiri D. (eds) MODSIM 2007 International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand, December 2007, pp. 74-80. ISBN : 978-0-9758400-4-7.
Stolk H. 2007. Risk Assessment for Flood Incident Management: Understanding and Application of Complex System Risk Assessment Models. RD Technical Report SC050028/SR5, Environment Agency, United Kingdom.
Stolk H.J. 2005. Emergent Models in Hierarchical and Distributed Simulation of Complex Systems. PhD Thesis, University of Queensland, Brisbane, Australia.
Stolk H., Gates K. & Hanan J. 2003. Discovery of Emergent Natural Laws by Hierarchical Multi-agent Systems. In IEEE/WIC International Conference on Intelligent Agent Technology, October 2003, pages 78-85, Halifax, Canada.
Stolk H., Gates K. & Hanan J. 2003. Emergent Models in Complex System Simulations of Genetic and Biochemical Networks. 11th International Conference on Intelligent Systems for Molecular Biology, Poster Presentation, Brisbane, Australia.
Stolk H.J. 1992. Parameter Optimisation of Control Systems using Learning Algorithms. MSc Thesis (in Dutch), Free University of Brussels, Brussels, Belgium.