How do you manage to introduce real aspects in your modelling?
I mean, MCDM does not exactly mean filling data in a model and running it. Real aspects of the problem must forcefully be considered. Perhaps the best way is illustrate this by some examples.
1-Precedence. Say that you have two projects A (build a government complex in an island in the middle of a city river), and B (Building a bridge connecting mainland with the island). How do you introduce the fact that to build the complex you first need to build the bridge? That is B > A.
2- A MCDM study calls for selecting either a tunnel, a bridge or cross the river using ferries. You only need one solution, that is, one project with ‘1’ score and the others with ‘0’ scores. How can you model this condition?
3- You are calling an international tender for supplying equipment for a hydro electric power plant. What provisions do you have in your model to allow joint ventures, which is two or more companies submitting together?
4- How do you consider in your modelling that if project C is selected, then project D can’t be selected?
5- How you tackle in your portfolio of projects the fact that there are already projects under different stages of construction, different completion dates, and you want to incorporate new projects? Of course you must indicate the model that projects underway must be finished; how do you model this?
6- In the last case you have a fixed annual budget to spend, however, each project has different annual percentages of completion and that annual budget can’t be surpassed
7 How do you manage to select projects in a portfolio when funds are not enough to execute them all? Say you need 32,457,617 Euros and only have 28,745,892 Euros.
8- - You are building a pulp mill in the north shore of a river, as well as developing crops agriculture, building two dams for water storage for irrigation purposes. You know that villages and towns downstream are arguing that they will receive contamination of from the plant, and more salinization as well as less water. How do you manage to satisfy all parts involved?
9- A multinational operate in farming, that is, different crops, cattle rising, fruit farms and juice concentrate plants in several countries. How do they manage to select which crop to cultivate in each country, how many hectares of each one, where to build the fruit farms and considering that to build the concentrate plants they need first to have fruit available nearby? How do you tackle this multiple scenarios problem?
10- A City Hall wants to execute a five year plan for the city infrastructure including sewage and LED street lighting, social undertakings such as developing community centers for youngster to get kids off the streets and building recreational parks, and similar projects for other areas. The whole plan has a healthy budget that will be handed over from the central government on an annual basis. How do you model your initial decision matrix for making the corresponding selection and be sure that each area receives the amounts of money it needs?
11- A DM delivers a ranking of projects for a City Hall ten years plan. Projects G-A-M-N-B-C-Z-Y are chosen and in this order of importance. The DM submits this plan to the Mayor and he asks ‘Where is project X that I promised to be built during my campaign for office? The DM explains that project X was originally considered but was not selected by the model. The Mayor says ‘I don’t really care what the mathematical model days, I want project X incorporated into the plan’, believe me that this happens and very often. How do you manage to get this project X in the final solution?
12- You have a budget for city infrastructure with money coming from Federal, Regional and Urban sources, each one for different purposes. The plan calls for renovating new bridges, demolishing old ones or building new ones. How do you manage to model this?
13- How do you consider correlation between criteria? Assume that you have determined that under certain conditions criterion ‘Snowfall intensity’ is highly correlated with ‘Speed’. How do you model that?
14- In an oil refinery, you have production, storage and distribution. How do you manage a model in order never run out of storage and never leave clients without oil products? It is relatively easy using linked algebraic equations. However, how do you introduce those equations in your model?
15- Once a MCDM is solved, the DM informs stakeholders which is the solution. How is the DM answer when one stakeholder asks how market share, demand and competition affect the selected alternative when the three of them vary simultaneously, with different positive and negative rates and independently?
16- How do you model a problem with say 245 alternatives and 891 criteria and with many of them related?
17- In an urban road project with three different alternatives, there is a survey amongst people affected by the projects. How do you introduce the statistical data in the initial matrix?
These are real problems and a MCDM method should be able to solve them.
If some practitioner ever found himself in any of these problems I believe that all RG community will be very grateful in knwing how it was solved and usinn which model
This is an exteeme dose of real world problems. If there is an answer, I would doubt it. Successfully meeting all demands may be impossible, there are tradeoffs in almost any decision. A heirachy of laws, regulations, guidelines and standards (BMPs, etc.) helps to set boundaries and guidelines to maintain a degree of quality and prevent blunders. Election, appointment and selecting competent individuals at all levels of governence and management also helps, with some allowance for public review and ability to comment in planning. It is very helpful in planning projects to have competent inventory, data collection, analysis, design, review by the appropriate technical, planning and managerial staff.
I cannot give you a name, but there is software to help plan and schedule projects of various types and complexities, some of which may have interdependence in each other. Environmental Impact Statements or similar documents should help with identifying and analyzing a reasonable number of alternatives for their effects to the environment, costs, social and any other conserns identified by decision maker or issues brought up in planning.
The value of having some boundaries is evident in setting a stage of operations and addressing many questions up front. Another aspect of planning involves having the possibility for civil or other actions to mitigate damages to those peoples or resources impacted. Planners, decision makers, land owners, managers and politicians are going to be more careful if their is some degree of liability if they make poor decisions or unreasonably impact their neighbors or resources.
At all levels, identifying, selecting, hiring, training of the appropriate staffing, whether it be civil engineers, biologists, soil scientists, economists, city or forest managers, agency heads, etc. will help insure access to needed information. The real world is often find a way to best utilize the staff available, offer training, set goals and determine objectives and priorities.
I was interrupted, continuing briefly only, the qestion is very complex. If you have circumstances with hundreds of alternatives or options, I would suggest trying to divide into parts or partition. Three to five alternatives including no action are desired. Hundreds of options may be analyzed with computer, but trying to explain so many is more than most publics and decision makers are willing to put up with. Someone must be responsible, or the result may be analysis and planning chaos. If there are "PET" projects of significant public support that are consistent with laws, regulations, etc., they probably should be considered and not easily discounted without reason. If mportant enough, lacking funds, perhaps contributions or cost sharing of resources can help address the need for resources. Voluntary services can sometimes be significant and could help avoid considering adding a tax to accomplish. In a coarse I took many years ago -- Koeptner-Trego decision making, a main emphasis was to write down listing of requirements and desired elements. Requirements being air, water, food, shelter, safety items and desires may be varied, such as a new car, a ski boat, a vacation trip, a recreation area, to things that borderline as requirements, such as adequate schooling for children, access to quality health and medical facilities, a well paying job, etc. Not everyone is going to see things the same way, and so there is almost an expectation that not everyone can be satisfied with decisions. When individuals are directly affected by decisions, it reduces the sting if there are some mitigation or compensation options.
In the first place we distinguish between the planning and decision-making processes which, although they may be complementary, have important differential characteristics.
One of them is precisely what refers to the information available: its quality and quantity, attributes that can only be assigned depending on each of the cases studied. That is, consider that each problem is different.
We deal with this problem using a hybrid model, which I can put at your disposal to carry out experiments on real life cases.
Dear Kevin
Could you be more explicit? What is it that you call 'Model parameters'?
I don't understand how you can use measured data to validate the problem if you do not know which the real answer is.. Yoy validate regarding what?
I agree that we have to introduce real data, no question about that, but my question is HOW do you introduce that data?
An example could be very illustrative.
Just pls. tell us how do you introduce that A > B? Unless you use Linear Programming of course.
Dear Melio
You are absolutely right of course. Planning and scheduling have nothing to do with MCDM, however, we may need information from the forrmer to feed the later
I am very interested in your hybrid model and I will take advantage of your offer,
My e-mail is: [email protected]
Thanks
Dear William
Your first writing in bold
Thank you so much for your answer.
This is an extreme dose of real world problems. If there is an answer, I would doubt it.
Yes, there is an answer.
As a matter of fact years ago I noticed that MCDM models did not represent reality since many aspects were ignored, and found that the best mathematical technique to model an actual situation and considering them - although not in a 100 % - was by using Linear Programming (LP). The lack of models for dealing with these actual issues prompted me to work during years on how to solve them, and in 2011 I was able to deliver a method called SIMUS (Sequential Interactive Modelling for Urban Systems), which is based in Linear Programming and Outranking. All the issues that I put in my questions can be solved by SIMUS, and I have examples for each one, which are freely available to anyone asking for them. Of course, there is software available which also takes care of sensitivity analysis. The software is free and fully operational and can handle hundreds of alternatives and criteria.
Successfully meeting all demands may be impossible, there are tradeoffs in almost any decision.
Naturally, not all demands might be satisfied, and the problem may be even unsolvable. The beauty of LP is that precisely it combines the multiples trade offs in such a way that a satisfactory solution is reached, because it is improbable that an optimal solution can be attained.
A hierachy of laws, regulations, guidelines and standards (BMPs, etc.) helps to set boundaries and guidelines to maintain a degree of quality and prevent blunders.
That is exactly what I said; nowadays MCDM models don’t consider boundaries which is not only a BPM, but in addition, perhaps more important, because nothing in our material world is unlimited. There are lower and upper limits for everything in funding, environmental contamination, rate of return, crime rate, water consumption, square metres for person in dwellings, people approval, available space, risk, you name it. These aspects are the raw material that most projects must comply to.
Election, appointment and selecting competent individuals at all levels of governence and management also helps, with some allowance for public review and ability to comment in planning. It is very helpful in planning projects to have competent inventory, data collection, analysis, design, review by the appropriate technical, planning and managerial staff.
No doubt about it
I cannot give you a name, but there is software to help plan and schedule projects of various types and complexities, some of which may have interdependence in each other. Environmental Impact Statements or similar documents should help with identifying and analyzing a reasonable number of alternatives for their effects to the environment, costs, social and any other concerns identified by decision maker or issues brought up in planning.
Well, I am a little confused here because you talk about planning and scheduling and I guess that you are referring to CPM and PERT methods with software such as Primavera and Project. It is precisely from these programs where the necessity of considering precedence between activities emerges, which was the first point in my presentation. However, I was talking about MCDM models and how to include those constraints as well as social and other concerns into the model. I don’t know any software, except SIMUS, that can handle this.
Regarding what you mention about Environmental Impacts Assessments it is fundamental and must be incorporated into de model.
The value of having some boundaries is evident in setting a stage of operations and addressing many questions up front.
I believe that is more than evident, however how may MCDM software use it?
Another aspect of planning involves having the possibility for civil or other actions to mitigate damages to those peoples or resources impacted.
This is a very interesting point that in all honesty I did not think about. Without too much thinking I suppose that remediation action as well as preventive actions can be established as criteria
Planners, decision makers, land owners, managers and politicians are going to be more careful if their is some degree of liability if they make poor decisions or unreasonably impact their neighbors or resources.
Exactly!
At all levels, identifying, selecting, hiring, training of the appropriate staffing, whether it be civil engineers, biologists, soil scientists, economists, city or forest managers, agency heads, etc. will help insure access to needed information. The real world is often find a way to best utilize the staff available, offer training, set goals and determine objectives and priorities.
I agree in a hundred per cent, however, if you need to select projects and you don’t reflect real issues in your modelling, you are wasting your time, and probably getting results that are not reliable.
Dear William
Your second writing in bold
I was interrupted, continuing briefly only, the qestion is very complex. If you have circumstances with hundreds of alternatives or options, I would suggest trying to divide into parts or partition.
Yes, you can do that, however, if alternatives are dependent you could be ignoring this fact.
Three to five alternatives including no action are desired.
Sorry, I don’t understand this. What do you mean by ‘action’
Hundreds of options may be analyzed with computer, but trying to explain so many is more than most publics and decision makers are willing to put up with.
This is true, for that reason you can decrease the number of options using for instance Principal Components Analysis.
Someone must be responsible, or the result may be analysis and planning chaos.
Naturally, a person must be responsible and this is the DM; the problem is that it is difficult with present day systems to prove that he or the stakeholders were wrong, mainly because the latter do not receive enough information on the consequences that certain future actions may have. This is why I pointed out the necessity to perform a rational sensitivity analysis, not as it done today
If there are "PET" projects of significant public support that are consistent with laws, regulations, etc., they probably should be considered and not easily discounted without reason. If important enough, lacking funds, perhaps contributions or cost sharing of resources can help address the need for resources.
Agreed, that this is the reason that the ‘PET’ project must be included; the problem is how you instruct the model that this project has to be selected; it is easily done in LP just stating as a restriction or criterion that PET > 0
If there is lack of funds, here is when the DM and stakeholders have to take decisions, even overriding what the mathematical model suggests.
Voluntary services can sometimes be significant and could help avoid considering adding a tax to accomplish. In a coarse I took many years ago -- Koeptner-Trego decision making, a main emphasis was to write down listing of requirements and desired elements.
Elemental!
However, KT refers to problem solving not to project selection, but of course, all the aspects they recommend to consider are valid also indecision-making
Requirements being air, water, food, shelter, safety items and desires may be varied, such as a new car, a ski boat, a vacation trip, a recreation area, to things that borderline as requirements, such as adequate schooling for children, access to quality health and medical facilities, a well paying job, etc. Not everyone is going to see things the same way, and so there is almost an expectation that not everyone can be satisfied with decisions.
In my opinion all those aspects that you mention are indeed requirements regardless their relative importance, and in addition they can be quantitative and qualitative. And truly, not everybody is going to see the same way, and this is one of the reasons by which criteria weights are useless to evaluate alternatives. The importance of each criterion depends on the values its respective vector t contains regarding all alternatives (performance values), but more important than that it depends of the existing discrimination between these values. Therefore, when there are many people affected by a criterion the best way is to obtain the performance values using Monte Carlo, or surveys or other statistical tool that best represent the average desire of people.
In addition, no limit must be established in the number of criteria; there must be as many as necessary. This is a fact that bounds the use of some software, and consequently by using them one can get results that have not considered reality, in other words, useless results.
When individuals are directly affected by decisions, it reduces the sting if there are some mitigation or compensation options.
Of course!
That is the reason that in some projects you must compensate people with new houses when they need to be moved, you have to compensate the environment when because of a projects forests are logged or flooded, you have to compensate down water villages because contamination produced by up waters industrial plants or water storage. All of these compensations involve more cost for the project and can be perfectly addressed through the corresponding criteria
Thanks Kevin for your explanation, which of course has a lot of sense.
Of course, you use an algebraic model with some unknown factors and you calibrate them using the true values, however in MCDM yout have neither an algebric model nor true values
Similarly, I used to perform calibration of a Gravity Model for determining flow between cities using real road data