1st International Congress on Environmental Modelling and Software - Lugano, Switzerland - June 2002
Keywords
graphical probability models, decision support, environmental modelling
Start Date
1-7-2002 12:00 AM
Abstract
We consider a novel approach to developing multi-objective environmental decision support applications. We propose using causal probabilistic networks (CPN) to subsume one or more engineering process models, together with CPN implementations of specific decisions and utility measures. So-called decision nodes are set to prior probability assignments of 1/N (where the particular decision has N possible values) and utility nodes are based upon the standard lottery principle give a value for that decision choice. The decisions can be so-called "one-shot" or sequential (time-dependent or at least seasonal). In the latter case maximization of the utility is calculated by gaming to optimize outcomes over a longer time period. The novelty of this approach stems from the embedding of the models into the CPN, and the potential for increasing the scope of decidability for environmental planners, and the entirely acceptable CPN approach of embedding evidential information in a heterogeneous fashion in addition to the modelling formulation.
Putting the Decision in Decision Support
We consider a novel approach to developing multi-objective environmental decision support applications. We propose using causal probabilistic networks (CPN) to subsume one or more engineering process models, together with CPN implementations of specific decisions and utility measures. So-called decision nodes are set to prior probability assignments of 1/N (where the particular decision has N possible values) and utility nodes are based upon the standard lottery principle give a value for that decision choice. The decisions can be so-called "one-shot" or sequential (time-dependent or at least seasonal). In the latter case maximization of the utility is calculated by gaming to optimize outcomes over a longer time period. The novelty of this approach stems from the embedding of the models into the CPN, and the potential for increasing the scope of decidability for environmental planners, and the entirely acceptable CPN approach of embedding evidential information in a heterogeneous fashion in addition to the modelling formulation.
