Biometrics & Biostatistics

Biography

Biography: Michael A Idowu

Abstract

Dynamic processes in complex systems may be profiled by measuring system properties over time. One way of capturing and representing such complex processes or phenomena is through ODE models of measured time series data. However, construction of ODE models purely from time series data is extremely difficult. First, the system targeted must be identified. Second, the parameters of the model must be estimated in a data consistent manner. Lastly, the constructed model must be capable of exact simulation of the measured historical data as though the constructed model was the means (source) of the acquired data. Hence, intelligent modelling of exact data may be a necessity in modelling systems that are not well-studied or well-known. The requirement to achieve the above-mentioned objectives within a short period of time, i.e., in order to cope with occasional or necessary demands of rapid data utilisation, makes both model construction and complex systems identification a modeller’s nightmare. In this presentation, a novel dynamic modelling technique (framework), invented and currently being further developed by the author, is proposed and presented as an effective computational method for reconstructing data-consistent ODE models, which adequately addresses the challenges of instantaneous systems identification and automated parameter estimation, under limited data and under-determined conditions. These dynamic modelling techniques (algorithms) enable data-consistent models of complex systems to be automatically constructed, with or without making a priori assumptions about the underlying network, which guarantees successful construction of feasible models in a matter of seconds. These claims are then justified with applications and examples.