A framework for inferring predictive distributions of rhino poaching events through causal modelling
Hildegarde Koen, J. P. de Villiers, Gregor Pavlin, Alta de Waal, Patrick de Oude, F. Mignet · International Conference on Information Fusion · 2014
Rhino poaching in South Africa is leading to a catastrophic reduction in the rhino population. In this paper a Bayesian network causal model is proposed to model the underlying (causal) relationships that lead to rhino poaching events. The model may be used to fuse a collection of heterogeneous information sources. If a game reserve is partitioned into several geographical areas or cells, the model may perform inference for each of these cells separately, and give a relative predictive distribution of poaching events over the game reserve. After an overview of the current problem definition and a brief overview of similar modelling approaches, the Bayesian network model is presented. The developed Bayesian network based model is an initial attempt at proposing a sensible modelling approach for this problem. Some of the complexities of the approach are discussed, before considering how the model may be validated at a later stage.