Use this URL to cite or link to this record in EThOS:
Title: Data-driven system dynamics modelling : model formulation and KPI prediction in data-rich environments
Author: Drobek, Marc
ISNI:       0000 0004 6425 2637
Awarding Body: Queen's University Belfast
Current Institution: Queen's University Belfast
Date of Award: 2017
Availability of Full Text:
Full text unavailable from EThOS.
Please contact the current institution’s library for further details.
System Dynamics (SD) is a key methodology for analysing complex, highly non-linear feedback systems. The SD modelling procedure is traditionally based on domain expert knowledge, manual modelling tasks and a parameter estimation and equation formulation process. These tasks are, however, heavily manual and complex since the required information was not expected to be available from written and numerical data sources. In recent years, we have seen an explosion in monitored and tracked system data that became known as the Big Data paradigm shift. This change has not yet found its way into the SD domain. Within this thesis, a novel data-driven SD modelling methodology for data-rich environments is proposed to address this paradigm shift. The research work carried out in this thesis exptores the potential of utilising massively available data sources for the SD modelling process. Based on these data sources, a modelling methodology (Fexda) is presented that supports the SD modeller in a systematic fashion whilst preserving the key principles of SD modelling. Unlike the traditional SD modelling, Fexda as a data-driven approach is highly sensitive to changes in the given data, which enables a continuous evolution ana optimisation of the computed SD models ana their parameters and equations. These contributions are based on advances in other domains, such as econometric modelling, data mining and machine learning, which are incorporated in a novel way for Fexda. A detailed evaluation of the proposed Fexda methodology is further provided against a business use-case scenario to demonstrate technical feasibility of the approach and to provide comparative results with traditional approaches. The evaluation clearly shows that Fexda can be employed to produce reliable and accurate SD models and provide insightful simulation results. The proposed Fexda methodology is the ground work towards data-driven SD modelling. A range of potential future research directions are proposed to further strengthen Fexda. The thesis concludes by presenting a revised version of the traditional information sources model that caters the reality of the Big Data paradigm shift.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available