FLASH SALE: 25% Off Certificates and Diplomas! Sale ends on Friday, 4th December 2020Claim My 25% Discount
We'll email you at these times to remind you to study
You can set up to 7 reminders per week
We'll email you at these times to remind you to study
So, it is really not possible for me to get into the I mean, intrinsic details of the different typesof I mean, approaches to multi criteria decision analysis. But, we will go through the broadand most important I mean, methods of MCDA and in the next lecture we would see anapplication of one method in an urban area.
(Refer Slide Time: 01:07)
So, the concepts that we are to cover today is we are going to talk about multi-criteriadecision analysis which shall include the definition and the application area and we would alsolook into the different popular MCDA approaches. I mean, there are lot of algorithms whichare available different approaches which are available as for MCDA analysis. Now, we willlook into the most popular of the MCDA approaches.
Now, we will also look into the different kind of software’s that are available for this popularMCDA techniques, we will look into what are endogenous variables. We will look how the
variable weightings are done and we will look into the structure of MCDA; Multi-CriteriaDecision Analysis.
(Refer Slide Time: 02:01)
So, let us begin with the first slide I mean talking about the definition and application. So,MCDA is I mean is used to resolve decision making problems; I mean, when you have I mean,a very complex problem and in when you have different variables in it. And, you would like tocome to a decision on the different kind of possible outcomes to choose which is the bestoutcome, then we generally used the MCDA kind of a approach to solve that.
So, these are formalised decision method which are analytical in nature. So, these aremathematical and I mean, this would give you a unique set of solution; I mean, anybody doesit they will arrive at the same set of solution. So, in a way it is I mean, very formalised method.So, it also has a very structured approach to formulating the data basis, to formulating the
entire methodology of applying this particular multi criteria; I mean, applications it employs avery structured approach to solving this kind of problems.
Now, we can create Decision Support System; so, in short we call it a DSS. So, we can Imean, use a decision support system it gives us I mean, multitude of factors and we can have amanagement tool based on these factors to come off with the best possible optimized result.So, I mean you can evaluate this I mean, multitude of your different input factors, you canoptimize them through some of these multi-criteria decision analytical tools and then you cancome to the most objective solution for this particular problem.
Now, there are different applications of your multi-criteria decision analysis. So, we are Imean just suggesting few, but you will see that there are a lot of other applications as well. So,you can see that there are applications in math’s, in management, in information technology,psychology, social sciences and economics.
So, there are different schools of thought and these approaches; I mean, are extensive andthere are more than about hundred different approaches of multi-criteria decision analysis. So,I mean it is really very difficult to be a champion in all of them. So, we identify I mean, thedifferent important multi criteria decision analysis techniques methods so, which we shalldiscuss shortly.
(Refer Slide Time: 05:00)
Now, the most frequently used multi-criteria approaches are the ELECTRE which is theacronym for your elimination Et Choix Traduisant la REalite So, it is a outranking approachthen we have a Multi-Attribute Utility Theory also known as comprehensively known asMAUT. So, it has aggregation approach then we have the ANP which is the AnalyticalNetwork Process which also is an aggregation approach.
So, then we have the I mean, MACBETH which is the Measuring Attractiveness byCategorical Based Evaluation, again this approach uses a aggregation method then we havethe analytical hierarchy process. So, in this also we aggregate I mean, we aggregate thedifferent weightages apportion to different criteria and inputs and it is aggregated to find outthe; I mean, final rankings of the different alternatives. So, then we have the TOPSIS which isthe acronym for tech technique of order of reference by similarity to ideal solutions. So, it is a
different approach where in we talk about the goal aspiration or the reference level I mean,approach.
So, then we have the PROMETHEE which is also used very commonly and it is the acronymfor Preference Ranking Organization Method for Enrichment Evaluation. So, of these, it isonly MAUT which uses only the quantitative indicator, but apart from MAUT all the othermethods, that we have talked about the approaches that we have talked about can use eitherqualitative, quantitative or even mixed indicators in the model.
(Refer Slide Time: 07:09)
So, it is only MAUT which uses only the quantitative indicator, we can also see the differentpopular approaches. So, I mean all the different approaches MCDA approaches that we hadtalked about; I mean, are they have two different macro phases. So, first is the construction
and the compilation. So, in this particular phase I mean, we evaluate the problem using a kindof a matrix or construct we construct a matrix.
So, we evaluate the different problem ah; the problem and then what we do is we havedifferent alternatives and the performance of these alternatives. So, that is I mean, included inthis particular matrix. So, and based on the criteria or the sub-criteria weightages we can haveindicators of assessments which are basically exogenous variables.
Now, the second phase; I mean, we were talking about two phases. So, first is we constructthe problem; I mean, a create evaluation matrix a based on criteria and sub-criteria of weights,the next one is the processing of data to evaluate this matrix used to evaluate the alternatives.So, we would have that the different alternative to a given problem like and then we need tofind out which is optimum or the best solution in a given scenario. So, I mean, on the basis ofthe objective that is to be reached I mean that is the endogenous variables, we do theprocessing of the data.
So, this method is similar for all the different MCDA method that is the processing of data inevaluation matrix. So, that is the endogenous variables, but when we are talking aboutexogenous variables that is the construction and the compilation then there would bevariability is in the different MCDA approaches that we have seen. So, there are differenttypology of indicators. So, I mean, in our earlier slide we have talked about I mean, thedifferent types of indicators. So, the; there could be qualitative indicators, you could havequantitative indicators and their also could be mixed indicators.
(Refer Slide Time: 09:35)
Now, the software that we have for this popular approaches. So, it is important that you knowabout these software’s so, that you can use them to process your data and I mean solve aproblem. So, we have the first bouquet of software is make it rational, we have expert choicefor AHP and super decision for ANP.
Now, this first bouquet of software it uses large number of criteria and sub criteria, but smallnumber of alternatives. Now, these three methods are participatory in nature. So, thatparticipatory approach is activated by taking feedback from the different stakeholders and Imean, preferably we try to I mean, group this stakeholders into different categories; I mean, itis putting this stakeholders in a very organized way.
Now, the second; I mean, method that is your MAUT we have the software which is known asright choice which is used to I mean, model a problem in a MAUT approach. So, in this we
have limited number of criteria, in the earlier three I mean methods or approaches we haveseen that we have large number of criteria and sub-criteria, but small number of alternatives.
But, in this right choice for MAU; I mean, MAUT approach we have limited criteria andsub-criteria and large number of alternative. So, it is just the reverse kind of a problem then wehad in the earlier three approaches. So, the nature of problem is difficult different, the problemthat you would be solving using MAUT approach.
Now, the in this method the participatory approach is limited with only very specializednumber of stakeholder. So, in our earlier three approaches ah; I mean, we had seen that wehave significant number of stakeholders and we organize the stakeholders into differentcategory. So, in this MAUT concept, there are limited number of participants and there arevery specialized stakeholders limited stakeholders for ah; I mean, evaluating your choices orgiving the weightages. So, we have very less number of stakeholders in the, who participate inthis MAUT approach.
The next approach is the ELECTRE. So, this can be I mean, you solved in this particularplatform which is known as Electre III-IV. So, in this, we have a limited number of criterialike the one we were take in talking about for the MAUT approach. So, we have limitedcriteria and sub-criteria and a small number of alternatives.
So, in this the n alternatives are valuate in relation to the objectives that is identified and in thiscase the participatory approach is not there which was there in the last two approaches thoughthe participatory approach was limited in case of MAUT. But, in case of electre thatparticipatory process is not activated..
The last one that is your TOPSIS, MACBETH and PROME you have these three software’s.For TOPSIS the software is called TOPSIS, MACBETH; I mean, you can solve or optimizethis; I mean, using this particular approach. You can optimize a problem by using the softwarecalled M-MACBETH and you can come to a solution using PROMETHEE approach using
SmartPickerPro. So, these are the three software’s in the, this last group of I mean, MCDAapproach.
So, in this there a large number of criteria like what we had seen earlier large number ofcriteria and sub-criteria, and also a large number of alternatives. So, both are large in this case.So, you can see the first point basically is a permutation and combination among this fourgroups of software. So, you now know that if you have large number of criteria, large numberof alternatives, small number of criteria, large number of alternatives your I mean and viceversa I mean, if you do a permutation and combination of these you can which I mean MCDAapproach you should choose to solve this basically.
So, in this these three methods in TOPSIS, MACBETH and PROMETHEE, you again I meanactivate the participatory approach that, there are significant number of stakeholders and theyare also organized in the category. So, there is a kind of I mean, this has been shadowed bythis particular object; so, this is PROMETHEE. So, this is a small typographical mistake thatyou have; I mean, which is has crop top in this particular slide. So, this smart picker pro isused for PROMETHEE the, or entire spelling is not seen in this particular slide.
(Refer Slide Time: 14:59)
So, you can see that we were talking about exogenous and endogenous variables. So, let ussee what is an endogenous variable. So, I mean, it is described as the type of decision makingproblem.
So, now, when we have this type of decision making problem; I mean, it is a description of theproblem and it identifies the main distinctive feature for a group of alternatives. So, you canhave a description problem and then there could be a sorting problem where in you wouldhave the definition of homogeneous group of alternatives by characteristics, then you can alsohave a ranking and choice problem. So, what we do is we rank the alternative from the best tothe worst. So, these are some of the type of decision making problems that we would comeacross.
Now, you can also have solution approach. So, in the solution approach, you have the firstmethod which is known as the full aggregation approach. So, in those case your score wouldbe evaluated for each criteria and these are then synthesized into a global score. Now, thisapproach would assume a compensable scores ah; I mean, that is if you have a bad score forone criteria it is compensated by a good score on another criteria.
So, the next one in the solution approach is the outranking approach. We have talked aboutthe full aggregation approach, now we shall talk about the outranking approach. So, in thecase if you have a bad score it may not be compensated by a better score like what we hadtalked about in our earlier full aggregation approach. So, in this case the order of option itmay be partial, because of the notion of incomparability is allowed in this particular approach.
Now, these two options may have same score, but their behavior may be different andtherefore, incomparable. We have this outranking approach I mean where in you see that thebad score may not get compensated by a better one. So, then we also have a goal aspiration orreference level approach in which a goal for each criteria is defined. We would define a criteriafor each of the goal sorry, goal for each of the criteria and then what we do is we choose theclosest options to ideal goal or the reference level would be identified in this particularapproach.
(Refer Slide Time: 17:56)
Let us see the other methods in your endogenous variables which are used. So, we shall lookinto the implementation pro procedure. So, in this we have the preference thresholds ah,indifference thresholds and veto thresholds. So, in this case your pair wise percentage degreecomparing the performance of n alternatives is made. And to find this preference level whatwe do is, we do an evaluation which must consider preference and indifference thresholds.
So, we are talking about preference thresholds and we are talking about indifferencethresholds in which a person or I mean, your decision making would be indifferent to the twooutcomes. So, we try to identify the thresholds and we also have veto thresholds which wouldbe I mean, which would negate those particular options or conditions which would negatethese options. So, on the basis of this thresholds; so, we can have positive, negative or
uni-criteria net and global flows which are created taking into account the weights that isattributed to each criteria.
Now, if we perform an action I mean, this action performs negatively I mean, if we have somekind of an action which is which performs negatively so, according to a single criteria. So, itmay also be included in a veto threshold that would definitely definitively exclude that optionfrom the final ranking. So, this is what we have in our preference threshold, indifferencethreshold and the veto threshold.
Now, we also might have a utility function which is an expression of the measure ofdesirability or preference of each alternatives with respect to the other alternatives ah. Now,there could be different criteria which are to be considered in function and when we are takingthis criteria for each of these criteria we have marginal utility which is determined asrepresenting partial contribution that each criteria belong brings to overall utility assessment.The global utility when we want to find out the global utility, it would be expressed by aglobal utility score.
So, generally this utility score varies, global utility score varies between 0 and 1. So, thisutility; global utility score are commonly calculated by a additive method or as a weighted sumbased on the weighted importance for each criteria or by a simple addition. Now, we can alsowork on pair wise pair wise comparison on a ratio scale. So, if you start reading into the AHP;so, I mean the analytical hierarchy process there we do a pair wise comparison.
So, we ask the stakeholders the experts regarding the importance of the different among thedifferent attributes or the different inputs. So, in this case what we do is we constructevaluation matrix which is also known as super matrix. So, the comparison of elementsincluded in super matrix which are these are organized into clusters of criteria, sub-criteria oralternative ah. It is performed by simultaneously come comparing two elements at a time. So,that is why we have this term called pair-wise comparison.
Because, we are simultaneously comparing only two terms at a time I mean, we also need totake into account any interdependencies between them. So, I mean, thus these
interdependencies is could be in terms of cluster criteria, could be in terms of alternatealternative cluster or there could be correlation between different clusters; so, I mean, basedon the influence between elements or clusters.
So, this super matrix that we are talking about is completed considering influence of each noteon others and it is expressed on a rational scale that is the satis fundamental scale. So,probably in our next lecturer we shall look into the satis scale. So, in case if there are nointerdependence between elements that are being compared, we can assign a vale of 0 in supermatrix.
So, we can also do a pair wise comparison on the ratio scale. So, I mean, we have talkedabout the ratio scale independence I mean, we have talked about the I mean; I mean, ratioscale we interdependencies then we can also do a pair wise comparison on the interval scale.So, in the interval scale what we do is we do a construction of evaluation matrixes which isalso known as the matrix of judgments.
So, the comparison is made between evaluation elements that is the alternatives and criteria, itis implemented by a pair wise comparison based on a semantic qualitative scale. So, a, I mean,traditionally the ranking or the I mean quantitative values is it ranges from 1 to 7 and thesevalues are included in matrix of judgments ah. Now, these judgments are relativeattractiveness of the criteria and the alternatives that we are talking about are also expressedby considering weights attributed to the each of these criteria.
Now, the next option the; next I mean procedure that we have is the ideal option andanti-ideal option. Now, this is an expression for each alternate of shortest distance to the idealor the virtual solution and the longest distance from the anti-ideal solution. So, it takes intoaccount the performance of the different alternatives which is referred to each criteria and toweight of each criteria. Now, the distance are expressed by calculating a distributivenormalization and an ideal normalization of recorded preference.
The next method is the output typology order of alternatives. So, in this particular methodwhat we do is we do a partial or complete order which is obtained by identifying or I mean,
jointing down the pair wise preference degrees or scores. So, what we do is we do asimultaneous consideration of positive and negative global performance flows which areevaluated for each alternative or by simply considering net flows, that make it possible tounderstand whether alternative is being deliberated obtain a higher rank or a minor rank, if twoor more alternatives are incomparable or equally valid.
We also do a partial and complete order which is obtained by expressing pairwise outrankingdegrees. So, in this we have degrees of reference which leads to a partial rank. So, it couldalso lead to a total rank of the alternatives and it is traditionally through expression of degreesof concordance or discordance I mean, that is according to the criteria that we areconsidering. So, we can do a partial or complete order obtained by expressing pairwiseoutranking degrees.
Now, we can also do a full order obtained by considering scores assigned to alternatives invarious ways by a we can do this by complex and general scores and we can and we can do ageneral approval or ordering of alternatives from the best to the worst in this particularmethod. So, we can also do a full order with a score closest to the desired objective.
So, in this what we can do is we can calculate the proximity coefficient for each alternativeand it is traditionally expressed as ranging from 0 and 1. So, I mean, you are if you have avalue of 1 that would express that it has the close closest proximity to the aim or the objective.
(Refer Slide Time: 27:20)
We have the next method which is the solution to the decision problem. So, in this case youmay have n categories of alternate alternative of equal score, but it may behave differently. So,you may have different types of behaviors for this n categories of equal scores, but these aredifferent alternatives. So, what we do is in this I mean, approach we considered the severaldifferent alternatives at the same time. And we identify the alternative with the highest globalscore by choosing the alternative which is I mean, which takes the highest score. So, I meanyour alternative closest to the I mean, with the closest score to the ideal solution is chosen.
So, we can also do a ideal normalization of recorded performance for the I mean, for thedifferent alternatives that you have. So, these are the different options when you want to do asolution to the decision problem.
(Refer Slide Time: 28:33)
Now, how do we do the weighing of the variables? So, this weighing all the variables aredepends on few factors, first is your variable variance in degree of the criteria how what is thevariance and we also look into interdependency. So, in our earlier slide we were talking aboutcorrelation. So, we talk about I mean, we should look into the interdependency of the criteriafor doing the weighing of the variables.
We shall also I mean, we have to look into the subjective preference of the decision makersbecause everything may not be very quantited quantifiable. So, you could have somequalitative aspects as well. So, there would be preference in terms of decision making,choosing some among the alternatives. So, we look into the subjective of preference and eitherthe decision makers or the stakeholders when we are I mean, putting or apportioning theweights.
Now, we may also have different weighing methods. So, the first weighing method this part isreally important that you should learn about the different weighing techniques and methods.So, the first method is the subjecting subjective weighing methods. So, these are weightswhich are directly assigned. So, these are simple multi attribute rating a technique which isalso abbreviated as smart there is SWING, SIMOS, pairwise comparison; I mean, AHP theseare some of the weighing methods.
So, I mean we also have objective weighing methods where in we talked about the entropymethod, there is a TOPSIS method of object weighing and combination of different weighingmethods.
(Refer Slide Time: 30:17)
I mean, if we see the structure of a MCDA solution. We would see that first we do an do aframework of the expected properties then what we do is, we calculate the overall index of
suitability and then we do in identification of the method which is best suited to resolving thedecision making problem.
(Refer Slide Time: 30:38)
So, you can look into the few books and references that we shall be talking about. So, ourrecapitulation of what we have done today is we had talked about the definition ofmulti-criteria decision analysis, we had seen the different application areas where this can beapplied, we had talked about the most popular MCDA approach and we had said that thereare more than about 100 approaches for MCDA.
So, we had looked into most of the popular approach and we also know that what kind of Imean, approach can be applied given your the different conditions in terms of your solutionssets or the number of variables in your particular problem or assignment. So, you can choosethe appropriate MCDA method, then we had also talked about the tools which you can use for
I mean, your data processing of your MCDA and we had talked about the endogenousvariables and we had talked about weighing of the variables. So, and finally, we have seenwhat is the structure of MCDA.
(Refer Slide Time: 31:46)
So, we have this following books first one is by Belton its I mean, Kluwer Academic Presspublication which is titled as Multi-Criteria Decision Analysis. The second book is by Figueira,so all these books are really interesting and good in this particular subject. Nijkamp book isvery good I mean it is the Multi-Criteria Decision Analysis for Land Use Management. So,those of few who are working with GIS or remote sensing you would find especially in yoururban applications, you would find this book to be really interesting ah; this is from the KluwerAcademic Publishers.
Now, we also have some reference to the research papers Saatys I mean, Thomas L Saatys Imean, who gave the AHP-ANP approach. So, he has written few interesting papers. So, this isone of them in from your operations research joiner ah. Roy B. Classement et so, he haswritten about your ELECTRE though this paper is not in English, but you will get thisparticular method in the books that we had talked about ah. There is paper by Dyer who hastalked about the multi attribute utility theory MAUT approach.
So, and then you have the analytical network process, ANP paper in ANP; so, this is also byThomas L. Saaty. Then Bana e Costa there he they have written and Vansnick they havewritten a paper on MACBETH. So, these are the different approaches another paper by Saatylooking into the AHP that is analytical hierarchy structures.
So, these are some of the research articles that you can refer to while you are I mean, studyingthis particular methods. So, this is a very interesting area and has I mean, wide applicability forI mean, coming to a conclusion regarding your urban issues and problems. So, this is a veryinteresting area to work in. So; I mean, the last one is the paper on PROMATHEE which is byBrans, this was published in the management of science general and thanks for I mean beingwith us in this particular lecture so.
Thank you so much.
Log in to save your progress and obtain a certificate in Alison’s free Urban Planning Criteria online course
Sign up to save your progress and obtain a certificate in Alison’s free Urban Planning Criteria online course
Please enter you email address and we will mail you a link to reset your password.