Reading about it, I noticed that there are many methods to resolve specific problems, but in general, the applications in MCDM, the AHP, ELECTRE and PROMETHEE methods are used more frequently.
Should we, as researchers, try to adapt the main methods or try to develop new ones, once if it will used only in few cases?
You forgot mentioning the class of value (utility) measurement methods. Actually the most flexible and robust MCDM methods.
Main MCDM methods have serious shortcomings. One of the most ineresting is a Rank Reversal (for example it occurs in AHP and ELECTRE). Let's we have a three objects (in other words: alternatives): A, B and C. If we use AHP or ELECTRE, we obtain the ranking. For example:
Priority of A > Priority of B > Priority of C
If we added a new object D, it would change the original rank. In the worst case we can obtain:
Priority of D > Priority of C > Priority of B > Priority of A
It is very hard to believe but the most important question is how much we can trust in final rankings. Which ranking is the proper ranking?
Priority of A > Priority of B > Priority of C (order without object D)
Priority of C > Priority of B > Priority of A (order with object D)
Yes, that is a Rank Reversal problem. I believe that we must find a verification method of different MCDM methods. In my opinion, it is the most important think to do.
The question is not how many methods we have but what is the real value of them.
Dear Melio Saenz, you have overwhelming right! A verification method of different MCDM methods is the most significance problem.
I would say that something that is for now missing in multicriteria decision aiding is a user-friendly software with a nice interface that would help anybody or any little company to use it in there everyday's life. If you consider a tool like diviz which is a nice software that computs hundreds of decision making methods, i had to spend many days to be able to apply it. Then maybe i was stupid.
You said (and you're right) that mostly, the used methods are ELECTRE, AHP and PROMETHEE. Personally talking about preference elicitation methods (not aggregation methods) i prefer using disaggregation approach. In aggregation approach you directly ask the DM about informations on the agregation parameters ("do you think the weight of this criteria should be higher than the weight of this one?"). I think that sometime (not always) DMs can have some difficulties to answer that kind of questions. In disaggregation approach you ask to the DM some informations about the output that he would like to get ("do you think that the alternative a should be ranked better than b?", "In what class do you think that the alternative a should go?"). ELECTRE and PROMETHEE are only aggregation methods (no confusion with "aggregation approach"!!!) while AHP makes both preference learning (with an aggregation approach) and aggregation method (an additive one). In that sense i think it is often hard to answer some the ahp questions such as "do you think the price of your next home is more important than its space?" ("well, it depends on the values that we have on these criteria because i couldn't buy a too expensive home nor live in a to small one"). But this is only my opinion.
To me, a very interesting preference learning method for additive aggregation is UTA.
Regards
Tom
There are many methods and many applications; the central point is the modeling process that involves the selection of an APPROPRIATE DECISION PROBLEMATIC.
Main problem is subjectivity, and then it does not matter the methiod you use, there is always subjective decisions. Consequently even using three or four different methods for a same problem you will get different results.
Even if different persons use the same method for a same problem most probably results are different becasue different decisions and preferences
The only method thst does not need ANY subjective opinión, perferences or assumption is Linear Programming. (LP)
SIMUS is a method based in LP and that in addition offers two solutions for a same problem and both giving the same ranking.
SIMUS software can be sent fully operational and FREE to anybody requesting it
For my money, every decision making method has some benefits and some drawbacks as well. Sometimes the best way is to combine two or more methods as an integrated framework to enrich our utilized methodology. As a researcher, it is so important how to integrate some MCDM methods to gain the best reliable results.
I think it is very crucial how to apply a certain decision making method for a specific problem. For most real cases, we need to use a combination of MCDM methods and compare the results with other methods. However, developing new INTEGRATED frameworks for real-life problems is more important.
I don't think that more applications are needed, there are hundreds if not thousands. In my opinión what is needed is a method that approximate the mathematical model to reality since most methods actually ignore actual facts and do not adequately represent a real scenario. In addition present day models are too subjective
- Badges
- Science topic
- Similar topics
- Indexes