| Conditional Outcome Preference Networks, also known as COP-nets, have been developed to graphically represent the model of a user's preferences over a set of possible outcomes. Typically, the number of elicited preferences upon which to construct a COP-net is limited. The structure of these partial preferences is then used to predict preferences over an entire set of possible outcomes.;The existing methodology for constructing a COP-net includes all possible outcomes and grows exponentially in the number of attributes that describe the outcomes, thus making the construction of the complete COP-net infeasible. In this thesis, a different approach for constructing COP-nets, using A* search, is introduced. With this new methodology, only outcomes that are relevant in determining preference over a given pair of outcomes are considered. Using this new approach, partial COP-nets can be constructed dynamically or on demand as opposed to the current process of generating the entire structure. Experimental results show that the new method yields enormous savings in time and memory requirements, and only a modest reduction in prediction accuracy. One such large example shows only a 5% decrease in the success rate, while reducing computation time from over 3.5 hours to just 2 seconds. |
