Hi Lavkush,
I guess you mean Spatial MCDA. To those mentioned by Stuart, you can also add gvSIG. It has a portable version which might help you to check its potential.
Hi, Lavkush,
in many Cases we need just tool for MCDA anslysis without GIS.
You can use for that DecernsMCDA (all the basic mcda methods + some advanced)
If U in fact are going to use GIS & MCDA with effective approach to creating spatial alternatives, U may try to use DecernsSDSS...
See my RG page and corresponding UsGuides/papers.
Write me , if...-)
=Boris
mr kumar patel you can use TOPSIS or AHP to solve your MCDA problem...
For academic use, you may download Visual PROMETHEE for free.
http://www.promethee-gaia.net
Visual PROMETHEE Academic Edition is an implementation of the PROMETHEE methods and is available for free for all non-profit research and education applications. Please visit www.promethee-gaia.net for download and don't hesitate to contact me for additional information.
Dear Lavkush Kumar Patel
There is the J-Electre, that implements several Electre methods: https://sourceforge.net/projects/j-electre/.
The Scikit-Criteria (a Python library) have some MCDA methods implemented: https://scikit-criteria.readthedocs.io/en/latest/index.html.
You can use R and the package RMCriteria. It implements the Promethee family and also the graphs.
Hi Lavkush Kumar Patel
There are the following plugins available in QGIS for spatial MCDA:
- Electre TRI: https://plugins.qgis.org/plugins/qgis_etri/
- GeoUmbriaSUIT (TOPSIS and DRSA): https://plugins.qgis.org/plugins/geoUmbriaSUIT/
- Vector MCDA (WSM, TOPSIS, Fuzzy MCDA, Concordance with discordance and concordance index, PROMETHEE and DRSA): https://plugins.qgis.org/plugins/VectorMCDA/
- Easy AHP: https://plugins.qgis.org/plugins/EasyAHP/
You can also perform the indirect integration with GIS by using other free tools. For instance, recently I used Visual PROMETHEE, Super Decisions, jMAF and MCDA ULaval, which are great MCDA softwares.
Try http://diviz.org/
It is a free framework with a lot of algorithm
Many thanks for the question and responses. We have an easy-to-use MS Excel based program for 10 different MCDM methods such as TOPSIS. It is free of cost. Please contact me at [email protected] if you are interested to have this program.
Selection of one of the Pareto-optimal (non-dominated solutions) involves weighting methods and selection/ranking methods. Our latest MS Excel program (version 2) has 8 weighting methods and 14 selection methods. User can use any of them as well as give his/her own weights for objectives. For more details, see Wang Z., Parhi S.S., Rangaiah G.P. and Jana A.K., Analysis of Weighting and Selection Methods for Pareto-Optimal Solutions of Multi-Objective Optimization in Chemical Engineering Applications, Industrial & Engineering Chemistry Research, vol. 59, pp. 14850-14867 (2020). Our user-friendly program is available free of charge to interested readers of this posting, by sending an email to [email protected].
Thank you very much Mr @Rangaiah,
This is both generous and interesting offer.
Best regards,
Alfredo
In my opinion, Excel is a wonderful tool for many different tasks, but not for finding the best option in a scenario. You need a dynamic algorithm for that, as found in MCDM methods
High Lavkush,
Please read a chapter of my PhD devoted to that topic:
Chapter Ranking of models given to my RG question
Greatings!
Alfredo
I have offered SIMUS many times
It is the easiest MCDM method to use, you only need to build your initial decision matrix, and it is one of the most powerful
Many thanks Nolberto Munier
I will include SIMUS on the list.
Any website where SIMUS can be downloaded?
Best regards,
Alfredo
I can recommend my team's library in Python, which is called pymcdm (main developers are Andrii Shekhovtsov and Bartłomiej Kizielewicz ). More info you can get in the project repository
https://gitlab.com/shekhand/mcda
You can install it by using in console:
>> pip install pymcdm
We still add new methods. Currently, there is:
- 15 MCDM methods;
- 10 weights methods;
- 8 normalization methods;
- 6 correlation coefficients (useful for analysis).
We have used it in many scientific papers, which can be found here: https://scholar.google.pl/citations?user=PAsiBvsAAAAJ&hl=pl
Thank you for the very useful source Nolberto Munier Wojciech Sałabun
Dear Alfredo
Thank you for your message
Regarding your second question, it could not have been formulated at a better time, since the SIMUS team, is at this very moment finishing the design of a Webpage for SIMUS.
During the last 10 years, we have tested SIMUS in probably more than 300 projects of the most diverse fields, from agriculture to healthcare, from thermal engineering to urban planning, to environment, to society, to road and railways transportation, to hydropower dams construction, to international help, etc.
This Webpage will be ready very shortly, and the user will find a lot of information about the method, with examples and with a blog where the reader may ask questions, require assistance to implement his/her project - in a way similar to RG - suggest new approaches, formulate new aspects and express his/her problems, even criticism, on SIMUS, which of course, will be answered by our team. In this section the user will be able to access the method, using the SIMUS zip file that will be attached.
The full version of SIMUS, allowing to work with hundreds of alternatives and criteria, has been, is, and always will be, free to everybody as well as the assistance through the blog.
Because its power, derived from Linear Programming, and due to using inequations instead of equations, as well as inputting in the same decision-making matrix, any mix of positive and negative quantitative and qualitative performance values, as well as those in binary format, and even algebraic formulas, the method is suited to model most of the characteristics of projects, such as precedences, ‘Yes’ or ‘No’ situations, or establishing special conditions, for instance, stating that the selected alternatives must incorporate a minimum of selected criteria, or reduce the number of initial indicators to a manageable final set, etc.
SIMUS does not need weights, thresholds, assumptions, preferences, or pair-wise comparisons, and then, the result is completely objective and a good base for the DM to take decisions. It is very easy to use since the DM only needs to build the initial decision matrix according to the problem
I am including the SIMUS zip file for anybody to download and share.
Dear Nolberto Munier,
I am afraid that SIMUS is not so ideal as you mention. I have worked with MCDA methods for a long time. Maybe, it is not so long as you, but I have an open mind, and I know that is not something like the gold method (which has only profits without negative). First of all, the SIMUS method uses the Simplex algorithm for Linear Programming. In the second one, the weights must be determined directly or indirectly. Let us suppose that we want to buy a car and we have only two criteria, i.e., C1 - price and C2 - quality. We have three cars A1 (x11, x12), A2 (x21, x22), A3 (x31, x32). How choose the best option when you do not know weights? Moreover, the second problem with nonmonotonic criteria, when we do not have linear programming? I think that SIMUS is a valuable method, but dear Nolberto, it seems that you forgot that decision-making is not only objective but also subjective. Very often, it depends on the human experience. In my opinion, the SIMUS is a nice method but with more limitations than these you present. It means it will be not the best one for each situation.
Dear Wang and Rangaiah, I like your Excel-based MCDM solutions and recommend them to everyone.
Dear Salabun's suggested MCDM solution sounds good, and I would recommend it to be Excel-based and user-friendly (my humble opinion).
On the other hand, the inclusion of normalization and ranking correlation types in the process seemed quite logical and innovative to me. By the way, my advice to Dear Salabun and his team is that they can offer a 'no-normalization' option for the MCDM procedure. It may be necessary for some problems that use unitless criteria. On the other hand, a similar procedure (no-norm.) can be tried for Entropy, SD, and CRITIC.
Finally, I have another interesting piece of advice for all MCDM software tools, which is the Rank Reversal problem. For example, does the MCDM method I use to have an 'RR' level problem? I want to learn this. What to do is simple. Add or remove alternatives. Then look and evaluate if the ranking has changed. The software can do this automatically. So he can determine the degree of RR
Dear Wojcieh
I am very grateful for your comments on SIMUS. Sincerely, you made my day!
Criticism on SIMUS is something that I was requesting for years in RG , and you are the first person that respond to my repeated requests. When the criticism comes from a well-known scientist like you, I feel rewarded. This is the reason I am happy with your comments.
Let me answer your comments:
1- First of all I never said that SIMUS is ideal, those are your words, and of course, it is not ideal.
2- ‘the gold method (which has only profits without negative)’. Could you please clarify this sentence of yours? I don’t know what you mean.
3- Effectively, SIMUS uses extensively the Simplex algorithm. You put it as it were something negative. Remember that it was chosen as one of the10 best algorithms in the XX Century, and remember that it is based on the Linear Programming concept that gave its creator, Kantorovich, the Noble Prize in Economics. What are your objections to using it?
Perhaps you ignore that since 1950 the Simplex was, and continues to be used for the largest oil refineries in the USA, for determining the amount of each refined product, subject to a lot of criteria, related to production, storage, transportation and demand.
4- Regarding weights, as you mention, I have never said that all criteria have the same importance, quite the opposite, it is unrealistic not to consider them.
The fact that LP or SIMUS don’t use weights does not mean that the relative importance of criteria is ignored.
If you know LP then, you are aware that its procedure is to investigate potential solutions in an iterative mode. There could be hundreds or thousands of iterations in a complex problem. For each iteration, LP computes the significance of ALL the criteria in a set, not only once as conventional methods do. The problem is that that significance is not computed using preferences or intuition, but using a mathematical procedure.
If you think about the existing methods, that determine a set of weights for criteria that holds constant, whatever the type of alternatives and applying to all of them, you will realize that it is obvious that this procedure is utterly wrong. Why?
Because a criterion may be very significant for one alternative but irrelevant to another, however, that fact is not taken into account. This is produced because weights at least in AHP and ANP, are 'computed' without
considering the alternatives they have to evaluate.
Therefore, to consider that weights have the same significance for whatever alternative is an illusion or a simplification. You can see here the advantage of LP, where the significance of criteria or ‘weights’ are computed for each alternative.
I ask you now: Which is more reasonable, the weights kept constant for all alternatives or they changing according to each one?
This LP system consists in selecting in each iteration, following the economic concept of cost of opportunity, the alternative that produces the maximum benefit or the minimum cost, and then, based on the aij of the corresponding column vector (for alternatives), performs a ratio analysis for each criterion to determine which is the criterion that influences the most in selecting the alternative to be deleted in order to keep constant the number of alternatives or dimensions.
5 – Sorry my friend, I don‘t understand this sentence of yours ‘Moreover, the second problem with nonmonotonic criteria, when we do not have linear programming?’
Criteria are in reality nonmonotonic, and what is the problem with that? Remember what I said above, alternatives in LP are solved one by one, independently, and by the way, this is the reason by which there is not a rank reversal in SIMUS.
I really don’t understand why you bring this subject; we are not talking about correlations.
6. No Wojcieh, I never forget that in MCDM there may be a very large component of subjectivity, and it would be naïve to ignore this fact. A good example is to select a restaurant for dinner, it is 100 % subjective. But is there subjectivity in location analysis, or working with the environment or with social issues? Yes, definitely there is, together with quantitative data, but not to be artificially generated, and altering data.
It appears that you suggest that SIMUS does not consider it, and in that, you are mistaken. SIMUS follows the bottom-up approach where you get an objective solution, and then the DM, using his/her knowledge, can accept it after his analysis, or modify the initial decision matrix, or correct it or simply reject the objective result.
It can detect for instance, that a certain issue has not been considered, either for omission or for whatever other reason. Say for instance that you are selecting among three lathes, and the objective result shows that lathe B is the best, followed by A and C.
However, there is something that makes the DM uneasy, because he is not comfortable with that selection. Then, he remembers that he has heard that this brand of lathe has had problems with its electronics. Therefore, he realized that a criterion is missing in the initial data, such as ‘Opinion of users of the three brands of lathes. Simply, he adds that criterion with the opinions of users and runs the software again. It may be that he gets the same ranking, or not.
What is paramount here, is that in building the initial matrix he could not know about this weakness of lathe B.
As a bottom line, the DM based on a solid result is able to analyze and decide.
Is it better or worse than altering at will real data with subjective weights?
Now, if you use entropy and statistical-based weights, that is different and valid.
Of course, you know that other than objective weights, subjective weights are useless to evaluate alternatives, because they do not have the capacity to do that. You know that said capacity is given by the discrimination of values in each criterion, something that is alien to subjective weights.
7- It would be very useful for me if you enumerate the other limitations of SIMUS. IT obviously must have limitations, however, I never said which are the ones you say I presented.
8 – I very often said that no MCDM is applicable to every situation, SIMUS included, but in all honesty, I don’t think that any other MCDM can match him in MODELLING and solving complex scenarios. If you want, we can perform a test on a complex problem comparing SIMUS with any other method, from the point of view of modelling. f you accept, please select the problem and the model.
Dear Mahmut
A no-normalization option? What if criteria have different units?
Entropy values are already normalized. This is the core of the method
There is too much discussion about the rank reversal problem, and many papers dealing with it on the Web
In addition, many researchers say for instance that there is no RR in ANP, when others disagree.
And how are you going to evaluate the alternative rankings after addition or deletion?
You can have a different ranking and it does not necessarily mean that there is RR, because it could be perfectly logical
Could you please explain how the software can automatically compute the degree of RR??
For my dissertation I created my own multi-objective decision analysis tool in Java. I used it to generate multi-dimensional tradespaces containing over 1 million design alternatives. If you are interested in the code please let me know.
Strongly recommended for AHP users https://bpmsg.com/
A real-world application https://www.springerprofessional.de/en/comparing-competitive-priorities-of-slow-fashion-and-fast-fashio/19581398
Dear Nolberto Munier
A&Q:
"A no-normalization option?"
-- The COMET method does not need normalization
"What if criteria have different units?"
-- The COMET method uses fuzzy logic and handling perfectly this issue
"Entropy values are already normalized. This is the core of the method
There is too much discussion about the rank reversal problem, and many papers dealing with it on the Web"
-- And so what? RR problem must be still discussed
"In addition, many researchers say for instance that there is no RR in ANP, when others disagree."
-- Some people are talking that Earth is flat and in my opinion, nevertheless, it is still geoid. The same thing with rank reversal and ANP. I don't see ANP which omitted RR. However, maybe therefore I am young :)
"And how are you going to evaluate the alternative rankings after addition or deletion?"
-- I do not understand the context
"You can have a different ranking and it does not necessarily mean that there is RR, because it could be perfectly logical"
-- Please give the EXAMPLE
"Could you please explain how the software can automatically compute the degree of RR??"
-- by using WS coefficients or other correlations measure.
Best Regards WS
Dear Wojciech
I did not know that the degree of RR could be measured.
Could you please explain it?
Thank you
Alfredo Erlwein Also we implement SIMUS in python https://scikit-criteria.readthedocs.io/
From this year on I gonna start updating the protect in frames os ~ 2 month.
The API is stable now, and the internal framework to create new models are sufficiently flexible to implement mostly any method in some hours (testing and documentation are slow)
If someone wants to collaborate or you need a more specific tutorial, please mail me.
Also, still no AHP. i not be sure how to bond the data structure of AHP with all the other methods.
Juan
You are right, AHP has a completely different structure, where the initial decision matrix in invented.
Right now, many methods use the output of the AHP first stage to weigh the criteria. Therefore all the drawbacks of AHP are transmitted to the other methods.
If you need, I have a lot of information that I can share with you
my email is [email protected]
Dear Rupak
Your first reference onPAPRIKA says:
'PAPRIKA method, which, as well as being scientifically valid,'.........
In my humble opinion, this is not true. Would you be interested in discussing it?
- Badges
- Science topic
- Similar topics
- Mathematics