A framework for modelling conflict-induced forced migration according to an agent-based approach
Thesis (PhD)--Stellenbosch University, 2019.
ENGLISH ABSTRACT: Over the course of the past decade, numerous calamities worldwide have led to phrases such as `refugee' and `undocumented migrant' becoming commonplace in the public discourse. Conflictinduced displacement and the various challenges it creates have received notable attention. A particular challenge posed by the management of sudden migration of large groups of people lies in a general inability to predict the scale and dynamics of such movement accurately. This problem is further complicated by the fact that data associated with such migration are largely incomplete or untrustworthy. Presently, there is a significant lack of data required to perform strategic, long-term planning related to current and future conflict crisis situations. One of the most fundamental challenges faced by researchers and humanitarian aid organisations when addressing forced displacement is an inability to predict the types of movement and the destinations of those who are forcibly displaced. The provision of a reasonably accurate estimation of the number of forcibly displaced people is potentially a critical input for the planning of logistics and management of structures supporting those eeing violence and persecution. A framework is proposed in this dissertation for assisting in the development and application of agent-based simulation models for predicting conflict-induced migration. The framework comprises five phases which encapsulate the formulation, conceptualisation and development of such a model, as well as the associated model execution and documentation. The purpose of the framework is to facilitate the design and development of agent-based models that incorporate the determinant factors of localised decision making and generate the resulting emergent large-scale movement patterns of forced migration. Collaboration with various subject matter experts throughout the development of this framework allowed for significant insight to be gained from the confluence of research in the fields of forced migration, computer simulation and human decision-making processes, which does not presently appear in the literature. The approach suggested for modelling human decision making is endorsed by knowledge gained from this research confluence, which has been corroborated by expert opinion. To the best of the knowledge of the consulted subject matter experts, no such framework encompassing such a wide variety of factors and implications pertaining to forced migration modelling in the presence of conflict presently exists and, as such, the research has sparked significant interest in the international research community. A concept demonstrator is furthermore developed according to the proposed framework in an attempt to demonstrate its usefulness and practicability in the context of conflict-induced migration in Syria. The model concept demonstrator is developed in the AnyLogic simulation software environment and allows for an animated output visualisation of the state of conflict pertaining to specified geographic and time scales, with super-imposed agent movement based on localised decision making when confronted with con ict. As per the framework, the model concept demonstrator is subjected to a number of traditional verification and validation techniques which include the calibration of parameters related to the modelling of con ict, the replication of visualised recorded data, the validation of the relevant agent aggregation, a thorough face validation and a parameter variation analysis which, ultimately, facilitates the implementation of a graphical user interface. The model concept demonstrator is thereby deemed capable of modelling specified scenarios of con ict-induced migration when equipped with the correct parameter values, owing to its exibility. The animation output also allows for easy interpretation of the model output, particularly by parties who are not necessarily scientifically trained. A framework of this nature naturally presents numerous avenues for future work. By employing machine learning and data mining tools, such follow-up work may, for example, ultimately lead to a framework for assisting in planning and the formulation of logistics strategies in the lieu of identifying required facilities and resources. Such an enhanced framework may prove invaluable in the accommodation of incoming refugees, internally displaced migrants and undocumented migrants in different areas, by predicting the population uctuations in affected areas during times of conflict, natural disaster, or other forced migration-causing events.
AFRIKAANSE OPSOMMING: Raadpleeg teks vir opsomming