Computational Recognition-Primed Decision Model Based on Temporal Data Mining Approach in a Multiagent Environment for Reservoir Flood Control Decision
In an emergency, decisions have to be made fast and accurate. Each decision has an influence to the safety of the public and properties. Due to the time pressure of the situation, a rapid decision model is required which will increase the speed of responding to an emergency situation. The decision...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | eng eng |
| Published: |
2004
|
| Subjects: | |
| Online Access: | https://etd.uum.edu.my/1487/1/NORITA_MD._NORWAWI.pdf https://etd.uum.edu.my/1487/2/1.NORITA_MD._NORWAWI.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In an emergency, decisions have to be made fast and accurate. Each decision has an influence to the safety of the public and properties. Due to the time pressure of
the situation, a rapid decision model is required which will increase the speed of responding to an emergency situation. The decision model must be able produce
accurate decisions due to the unpredictability and uncertainty of the situation that develops. This study is an initiative towards developing a computational model for
emergency decision-making. The characteristics of an emergency environment resembles naturalistic decision-making environment. Among the properties of this environment is time pressure, urgent, unpredictable, high uncertainty, high stakes, usually involved multiple players and experienced decision-maker. In a situation such as this, experience and able to recognize a similar situation with the past is essential. This recognitional strategy helps reduce time taken in making decisions by comparing
to previous decision patterns. Gary Klein introduced such a model called Recognition-prime decision (RPD) to describe experienced decision makers thinking processes. For a computer system, the computational model should be able to learn from experience and recognize a similar situation. The 'learning from experience' requirement resembles a computational intelligence procedure called data mining. It
can provide an autonomous decision-making capability that can facilitate shorter response time in decision-making. Emergency situation has special characteristics where events are usually an effect of a cause after some considerable delay. Hence the situation is time dependent and strictly time: ordered. Therefore the data mining
approach need to be able to handle this time factor.
This thesis, explored the feasibility of using temporal data mining approach to support computational RPD model to be used in a naturalistic situation. Flood emergency is taken as domain to be studied due to its common occurrences in Malaysia. Reservoir flood control at Timah Tasoh dam in the State of Perlis is taken as case to test the model developed. Real operation data were used to validate the
model. A multiagent system was also designed, prototype implemented and tested to also provide autonomous and proactive capability to the emergency warning system. Agent based approach and temporal data mining provide faster response to impending emergency situation. Performance of the model was measured against real operation data from 1998 to 2002 and was found to predict with more that 90% accuracy with less than 10% false alarm. |
|---|