AICME II abstracts
General
A Multiattribute Decision System for Selection of Environmental Restoration Strategies Sixto R´ıos-Insua1 , Antonio Jim´enez2 and Alfonso Mateos3 . The selection of the most preferred remedial strategy in complex environmental problems should be based on all the relevant information, being the ultimate objective to minimize the global impact on the scenario. To afford such complex problems under the presence of multiple conflicting objectives, we have developed a decision support system based on the decision analysis cycle that allows for considering all their stages [1]. Initially, it is possible to model an objectives hierarchy with attributes associated to the lowest-level objectives Xi , i = 1, ..., n, which permits to provide a basis upon which to develop and appraise screening criteria. Thus, the consequence of each possible decision strategy Sj can be de j j j scribed, under certainty, by a vector x = x1 , ..., xn , where xji is the level consequence of attribute Xi for Sj . Under uncertainty, ithe h i ofheach strati h L U L U U egy is given by a vector of ranges x , x = x1j , x1j , ..., xL , x nj nj , where L (U ) means Lower(U pper) endpoints of the uncertain consequence [2]. Next, the system allows the decision-maker (DM) to quantify the preferences over the different attribute ranges by assessing imprecise single attribute utility functions, as well as the relative importance of objectives in the hierarchy through/by means of imprecise weights. The strategies evaluation is performed by means of an additive utility function, whose Pn j j form is u S = i=1 ki ui xi . In the case under uncertainty, utilities intervals are also provided by means of
hP n
L i=1 ki ui
P n L
xij ,
U i=1 ki ui
xU ij
i
,
General
AICME II abstracts
which provides us useful information about the ranking robustness. When no weights have been assigned the system uses interactive simulated annealing to aid the DM to reach the best or a satisfactory strategy. As an essential complement of any quantitative model, the system includes several sensitivity analyses that can be an important aid to study the robustness of the final ranking of strategies. First, it provides several displays among which there is one that permits to watch the strategies ranking from their precise values and includes their associated utility intervals. It also computes and displays the weight stability intervals for each one of the objectives of the hierarchy and provides a stacked bar rankings, which shows the average utility assigned to each strategy by breaking down the contribution in the utility bar for the different attributes. The system also exploits the imprecise information on weights, utilities and consequences by computing the non-dominated and potentially optimal strategies by solving certain optimization problems [3]. Finally, it is possible to performs simulation techniques which allows for simultaneous changes on weights and generate results that can easily statistically analyzed to provide more insights into the model recommendation. Acknowledgements: Supported by the Ministry of Science and Technology project DPI2001-3731.
References [1] Jim´enez, A., S. R´ıos-Insua & A. Mateos, 2003, A Decision Support System for Multiattribute Utility Evaluation based on Imprecise Assignments, Decision Support Systems (to appear) [2] Mateos, A., S. R´ıos-Insua & E. Gallego, 2001, Postoptimal Analysis in a Multi-Attribute Decision Model for Restoring Contaminated Aquatic Ecosystems, J. of the Operational Research Society 52, 1-12.
1 Department of Artificial Intelligence, Madrid Technical University, Campus de Montegancedo s/n, Boadilla del Monte, 28660 Madrid, Spain (e-mail:
[email protected]). 2 (e-mail:
[email protected]). 3 (e-mail:
[email protected]).
[3] Mateos, A., A. Jim´enez & S. R´ıos-Insua, 2003, Solving Dominance and Potential Optimality in Imprecise Multi-Attribute Additive Problems, Reliability Eng. and System Safety 79, 253-262.
Gen-Rio-a
Gen-Rio-b
