Technique of order preference by similarity to ideal solution (TOPSIS) is a useful technique for situation like this. Please try its fuzzy version.

Best

Have you tried to sort the possible solutions according to the pleasure (or minimal displeasure) they give you? See: Cabanac M. (1992). Pleasure: the common currency. J. theoret. Biol. 155: 173-200.

Good luck!

Some useful papers:

http://www.ips.iit.bas.bg/D_Borissova/AMO_2008_v10(1)_81-93.pdf

http://waset.org/Publications/investments-attractiveness-via-combinatorial-optimization-ranking/17005

Thanks for your answers.

The main goal I want to achieve is the following:

I want to derive an importance ranking of my activities based on the selection of characteristics (one characteristic for each of the 16 aspects). This can be done by a (static) matrix, which adds up the influences on each activity according to the selection of characteristics. Challenge is, that the values in this matrix need to allow different rankings based on different characteristic profiles.

Thais Matrix I want to develop - with ANP, indirect influence analysis or whatsoever. ANP therefore has the disadvantage, that it gives a huge amount of pairwise comparisons which is too much effort for that case. I did it in Excel before with two influence matrices, but of course it is highly dependant on the scale used.

I don't know whether you already get a ranking among 21 activities. I am not really understand "useful results". Do you mean the results are not accuracy? or the results cannot give you the corresponding means? If in the case, i suggest you use fuzzy numbers in the beginning to express linguistic terms instead of crisp numbers.

Or you can try a compromise solution which to find out the closet idea alternative. (VIKOR)

Sanayei, Amir, S. Farid Mousavi, and A. Yazdankhah. “Group Decision Making Process for Supplier Selection with VIKOR Under Fuzzy Environment.” Expert Systems with Applications 37, no. 1 (January 2010): 24–30. doi:10.1016/j.eswa.2009.04.063.

hope it will hope you.

Another useful approach to your problem:

A Web Application for Group Decision-Making based on Combinatorial Optimization

I believe that the reason you can't get "useful" (rational) results is because "aspects"(importance criteria) "influence each other" (ie they should be "independent"). Try a different modelling of this ranking problem (different set of criteria).

Also: AHP and ANP are not particularly useful methods for evaluation of such a large number (21) of alternatives -especially on 16 given criteria. I would recommend a multicriteria analysis method such as UTA http://www.unipi.gr/faculty/ysiskos/files/paper3.pdf

This paper can help you as an example.

Azizi, M. and Modarres, M., (2011), A decision making model for investment and development of construction panels, Journal of Forestry Research, 22 (2), 301-310.

My experience with ANP on superdecisions suggests that it is necessary to have bidirectional relationship between clusters. otherwise, it will not give good results and probably in that case AHP works better than ANP.

Here are some papers for your consideration:

http://ijbssnet.com/journals/Vol_3_No_11_June_2012/22.pdf

http://www.mii.lt/informatica/pdf/INFO870.pdf

Prof Saty's paper remains a good source: http://link.springer.com/article/10.1007%2Fs11518-006-0151-5

This is another reference in Super Decision Software for your consideration.

All pairwise MCDM methods are very cumbersome when there are more than 6/7 criteria. An alternative is to combine a pairwise method (just to get weights) and then use a decision matrix with a "clever" aggregation process. E.g.:

R. A. Ribeiro, A.M. Moreira, P. V.D. Broek, A. Pimentel. Hybrid Assessment Method for Software Engineering Decisions. Decision Support Systems 51 (2011): 208-219. DOI: 10.1016/j.dss.2010.12.009

THanks for all the papers and helpful hints. I will go through them and I am sure to find a solution for my problem - it is actually not that difficult I think, the challenges that makes it inconvenient is the number of elements related to each other. ANP is definetely impracticable due to the amount of pairwise comparisons that would be necesasry.

I included an overview of the Problem (as modeled for ANP).

Be selection of one characteristic for each of the 16 features (blue), a rankingfor the 21 planning activities (green) should result. Since it is sufficient, to do that quasi static I want a matrix that gives me the influence of each characteristic on the planning activity, so I can simply sum the numbers for each activity for each selected characteristic.

Maybe an indirect influence analysis is even sufficient (does anyone know a free template or software for that?).

Dear Flo Mue,

You have to consider the absolute measurement mode of the ANP, other way the number of comparisons among the alternatives will be huge.

What does absolute measurement mode mean?

As far as I know ANP, absolute measures can be used to evaluate the likelihood/importance in the pariwise comparison, how could they be avoided?

I'm talking about create scales of measurment (absolute measurment) one for each terminal criterion, and then rating the alternatives one by one against those scales (instead of compare the alternatives among themselves).

This will return you for each alternative a "behaviour profile", which normally is a very useful information in planning problems.

Have a look at this paper (www.nit.eu/czasopisma/JTIT/2010/4se/15.pdf‎), you may find an answer to your question.

I think that you should check this presentation: http://www.superdecisions.com/tutorial-4-changing-from-ahp-to-anp-thinking/

that should help on polishing a little bit the model and maybe then work with ratings due to the large number of alternatives to be evaluated. You can see the attached file on how to work with ratings.

You will be not able to solve your problems with ANP as ANP is making fundamental errors, namely:

-uses weights meaning a dual situation of normalization and importance

- is a victim of the Condorcet - Arrow Paradox by comparing two by two

- consequently can not be used for large matrices

- additional remarks of Chakraborty:

computional time very high

simplicity very critical

mathematical calculation involved: maximum

stability:poor

information type: mixed

Use rather other methods. Therefore see my contributions to Reseach Gate

Dear Mr. Brauers,

Seems to me you are mixing concepts applicable to ordinal and cardinal numbers. For instance, Condorcet and Arrow’s Paradox, (which is really Arrow´s impossibility theorem), is intended for ordinal numbers not cardinals, in fact there is a very nice demonstration (from Saaty’s book Group Decision Making, RWS, 2010) that, when dealing with cardinal numbers the impossibility theorem becomes possible (is superseded). Is very important to understand the difference between this two set of numbers, since the scales of measurement are very different. AHP/ANP works with absolute scales of measurement, a special case of proportional or ratio scales, it is dimensionless and absolute (no zero definition needed). About the comment on to avoid on large matrices (comparing two by two on large matrices), that’s correct, and that’s where the knowledge of the problem is important to model it correctly decomposing into smaller pieces, and integrating back by the synthesis process.

AHP/ANP has been used to successfully resolve significant real life problems by very knowledgeable professionals, and also by simple citizens with reduced academic background but lots of understanding in the problem being modeled and even better, it has been tested on those problems where “correct” solutions exist, with great results if the participants really know about the problem they’re trying to solve.

Your comment seems from someone with little experience using AHP/ANP on real life situations. I invite you to expand your knowledge. Most critics against AHP/ANP have all been academically responded. But still, individuals have the right for their personal preferences… as far as recognized as such.

Hi everyone! hope you are doing well.

I would like to identify and rank barriers to the development of renewable energy sources using Super Decision. I have 13 criteria and 3 alternatives. Is there any limitation on the number of criteria in this software?

I can only enter 12 criteria?

Please give me a hand to find solution.

Thanks in advance