Fenton and Neil illustrated the use of MCDA in conjunction with Bayesian Belief Networks (used to produce the  Ratings on 3 of the 5 Attributes) for choosing a travel mode + leaving time Option for a journey to the airport. The flight leaves at 8.45 am - Comfort
- Cost
- Journey Time
- Start Time
- Airport Wait
- Car leaving 4-5 am
- Car leaving 5-6 am
- Car leaving 6-7 am
- Car leaving 7-8 am
- Taxi leaving 5-6 am
- Taxi leaving 6-7 am
- Taxi leaving 7-8 am
- Train leaving 5-6 am
- Train leaving 6-7 am
- Train leaving 7-8 am
Supplied by authors. Note that Waiting time is non-linear with 30-60 minutes rated 1, 60-90 minutes 0.7, 90-120 minutes 0.5 and more than 120 minutes 0.3; but 15-30 minutes is rated worst of all at 0.1 and less than 15 minutes is labelled 'reject' on grounds that the flight would be missed. In our Annalisa we rate >15 minutes at 0 and show that options involving it do not emerge as optimal anyway, but the correct way to deal with such threshold constraints is to eliminate Options involving them. The high weight attached to Start Time leads to the 3 Options leaving at 6-7 am emerging in the first 3 places. In slide 2 we show that even higher weight would lead to the very tight Train + 7-8 leaving Option N Fenton, M Neil (2001) 'Making decisions: using Bayesian nets and MCDA' Knowledge-Based Systems 14: 307-325 The second illustration in the paper is a more complex one relating to the licencing of computer software for nuclear power plants. download alt file here
|
