Exploiting Decision and Policy Models More Effectively Through Intelligent Search, Data Mining, and Visualization
Abstract
Candle-lighting analysis encompasses a series of automated techniques for supporting intelligent examination of decision and policy models. Such examinations are well-suited for model validation and for post-evaluation analysis, and can be particularly useful for providing decision makers with insight. The primary alternative to candle-lighting analysis is manually-directed what-if questioning. Such what-if analysis is a cumbersome process, appropriate mainly for small-scale models; it is a process forever at risk of being undermined by human biases and cognitive limitations. Candle-lighting analysis concepts hold the promise of greatly enhancing the value and effective use of management science models in the U.S. Army, and DoD generally. Essential to the concept of candle-lighting analysis (CLA) is the creation of multiple databases containing information gleaned from particular formal or mathematical models using intelligent heuristics (including, but not limited to, genetic algorithms). CIA treats models as complex, hyperdimensional structures in decision space. The resulting (hyperdimensional) surfaces need to be explored intelligently and mapped for human decision makers. These explorations yield large data sets, which are stored in a database environment. The databases are then mined in order to give decision makers insights in the problem at hand, and to suggest promising policy options to them.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 04, 2001
- Accession Number
- ADA391258
Entities
People
- Steven O. Kimbrough
Organizations
- University of Pennsylvania