A Module to Estimate Numerical Values of Hidden Variables for Expert Systems.
Abstract
In the area of strategic decision-making, the objective often is to achieve one's own goals and to prevent the achievement of the adversaries' goal. To do so, the decision-maker needs to know, as precisely as possible, the values of the relevant variables at various times. Some of these variables, the open variables, are readily measurable at any time. Others, the hidden variables, can be measured only at certain times, either intermittently or periodically. The authors have implemented a module that can act as a decision-support tool for a variety of expert systems in need of estimates of hidden variables values at any desired time. The estimation is based on generalized production rules expressing stochastic, causal relations between open and hidden variables. The quality of the estimates improves through a multi-level learning process as both the number and the quality of the rules increase. The modularity of these causal relations make incremental expansion and conflict resolution natural and easy. Restricting the set and the domain of pattern formation rules to a reasonable size makes the system effective and efficient. Finally, the system can be easily employed for distributed database applications. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1981
- Accession Number
- ADA110256
Entities
People
- Han Yong You
- John E. Brown
- Nicholas V. Findler
- Ron Lo
Organizations
- University at Buffalo