Evaluating Decision Trees under Different Criteria

Danielson, M. & Ekenberg, L. ORCID: https://orcid.org/0000-0002-0665-1889 (1996). Evaluating Decision Trees under Different Criteria. IIASA Working Paper. IIASA, Laxenburg, Austria: WP-96-148

Preview

Text WP-96-148.pdf Download (465kB) | Preview

Abstract

Based on our earlier results in decision theory, we demonstrate how decision trees can be integrated into a general framework for analysing decision situations with respect to different criteria, and suggest an evaluation rule taking into account all strategies, criteria, probabilities and utilities involved in the situations under consideration. A significant property of the framework is that it admits the representation of imprecise information at all stages. This information is modelled in sets of measures constrained by interval estimates. The strategies are then evaluated relative to different decision rules, e.g., a set of generalisations of the principle of admissibility. Decision situations are evaluated using fast algorithms developed particularly for solving these kinds of problems. The presented framework has been developed and used within a large-scale evaluation project at the Swedish National Rail Administration.

Actions (login required)

View Item