Dear Tan

It is difficult to evaluate which method to use, because it depends on different circumstances, the nature of problem and the kind of data you have.

May I suggest you to read the paper written by Prof. Alessio Ishizaka?

it is very illustrative, and I think it can help you.

Alessio Ishizaka. Multi-Criteria Decision Analysis Methods and Software

http://onlinelibrary.wiley.com/book/10.1002/9781118644898

Dear Dr Nolberto,

Thanks for your suggestion.

I will read the paper as well.

I have a big question in my head,

if I choose either one method, lets say I choose the F-TOPSIS, do I have any evaluation method to prove my choice is valid for my recruitment problem ?

Just wondering if Dr. can suggest me any related paper that mentioned about evaluation method on the FMCDM. =)

Thank you once again, Dr Nolberto.

Dear Tan

All depends of what you call 'valid'. If the method you use, any method or model, satisfies the Decision Makers and or stakeholders, I believe that it is enough prove or solution. To assert that it is the best, well nobody can, because you will never know what would be the results if you recruit other candidates

Regarding your request I would recommend reading.

Ozlem Senvar, Gülfem Tuzkaya and Cengiz Kahraman

Multi Criteria Supplier Selection Using Fuzzy PROMETHEE Method

They are in RG.

Good luck

Nolberto

Dear Tan

Trying to evaluate MCDM methods is an old aspiration for scholars, however as far as I know there is no still a procedure for that. The main reason is in my opinion the fact that we do not know the result, by which can we evaluate the reliability of a method. For instance, if you have an equation such as y = a + x2, you have the certainty that whatever value you have for x, the resulting result for ‘y’ will be correct. In MCDM we do not have such a formula, and as a matter of fact if we could have  a formula to determine the best ranking, we could use it and no MCDM model would be necessary……

In addition, remember that the different MCDM models, even based on sound mathematical procedures (albeit not everybody agrees with it), use different procedures mostly based on personal preferences, estimated thresholds, assumed distances, fabricated weights, etc. This is the reason by which a same problem, solved by different methods often yield different results. But how do we know which is the ‘best’?

Attempts have been made by different scholars, mainly pioneered by Triantaphyllou and Wang to find an answer to this query.

 If your are interested I would suggest reading this very interesting and enjoyable paper  by

 Xiaoting Wang1 and Evangelos Triantaphyllou.

‘ Ranking Irregularities when Evaluating Alternatives by Using Some Multi-Criteria Decision Analysis Methods’

You can easily access to it in Internet entering with this title.

In my opinion I believe that we could use some sort of benchmark model such as the ‘Simple Additive Weighting’ (SAW) method, one of the first to be developed and indeed very simple. However, I also believe that it is too simple to be compared with models that can treat complicated and complex scenarios.

It is also my opinion that we can measure a model based on its ability to replicate, although in part, a real scenario and especially where subjectivity is reduced to a minimum. However, probably the most important issue to select a model is to determine which is the model that best adapts to your needs

I hope this answer of mine can help you, but do not hesitate in contact me again if you wish to further investigate this issue.

Nolberto

I suggest to stay away from Fuzzy AHP since there is a paper by Saaty that suggests there a lot of subjectivity in AHP by itself so adding fuzzy makes the results unreliable, if the subject of MCDM is of interest to you, I would suggest the classical methods such as fuzzy weighted average, AHP, ANP, ELECTRE, Multi Attribute Utility theory, VIKOR, or as professor suggested SAW or PROMETHE, which is straight forward calculation, you only introduce fuzzy if the data you are entering is subjective in nature using fuzzy membership functions you go from linguistic to the fuzzy domain, hope this helps

Dear Tan

Comparing of different MCDA methods is the multiple criteria problem in itself, for which you need some kind of the best method, defining such method was the original aim ... Professor Nolberto Munier is wright. 

All of that is quite situational. I believe that comprehensiveness and easy-to-implement are the suited criteria in your case. No trade-off will arise, there is a dominant solution: AHP.

AHP allows to combine different types of criteria, based on qualitative judgments and on quantitative data. Uncertainty is already incorporated, though there will be no gain in fuzzifying an AHP model - more effort, same or less precision.

Best regards

Thanks for the suggestion. I will look forward for it. 

Really appreciated

Dear Tan

In this case I would recommend AHP, because this is clearly a matter of preferences and most especially because the recruiter will be the beneficiary if his decision, so, he is not taking a decision for others but for himself of for his own company, and thus he knows better than anybody else what criteria are more important for his company.

Then I agree with Professor  Kisly that AHP is your best option.

Also you can do something better, which is determine weights by AHP and then apply TOPSIS

Dear Dr Nolberto Munier, 

I will look forward for determine weights by AHP and apply TOPSIS for the decision making on alternatives.

I have few questions, can I ask them here ?

1) Is FMCDM like F-AHP and F-TOPSIS using rule-based ? Meaning to say, I need to construct the rule in order for the system to make decision. 

OR 

they are NOT using rule-based. And what I need is to determine the weight and construct the membership function for each of the linguistic variables (criteria) that I have. 

2) How to determine the weights? (by human resource experts) ?

3) How to construct the membership function for linguistic variables, e.g. my first linguistic variable is "RESPONSIBLE", value like very not related, not related, neutral, related, very not related.

Thanks for your attention. I am really appreciated it.

Dear Tan

Tan. I will look forward for determine weights by AHP and apply TOPSIS for the decision making on alternatives.

NM. Fine, I think it is your best option

Tan. I have few questions, can I ask them here

Tan. 1) Is FMCDM like F-AHP and F-TOPSIS using rule-based ? Meaning to say, I need to construct the rule in order for the system to make decision. 

NM.I imagine that FMCDM means Fuzzy multicriteria, however, whatever the model you choose and if you apply fuzzy logic, obviously you have to follow those rules

Tan. OR 

they are NOT using rule-based. And what I need is to determine the weight and construct the membership function for each of the linguistic variables (criteria) that I have. 

NM. Well, you need the rules to build your memeberships function, and when you fet the result, defuzzify it and you get the cdrip number that you can put in any model.

Tan. 2) How to determine the weights? (by human resource experts) ?

NM. That is the reason that I recommended using AHP. This model has two stages; in the first one you determine criteria weights by pair-wise comparing criteria against the main goal. In the second step you pair wise compare alternative against each criteria. Once you get it you multiply each one by the corresponding criterion weight. 

Tan. 3) How to construct the membership function for linguistic variables, e.g. my first linguistic variable is "RESPONSIBLE", value like very not related, not related, neutral, related, very not related.

NM. You have to decide the format of the membership function, that’s triangular, trapezoidal or normal