Professional Documents
Culture Documents
DSS/GIS
Objectives
This lecture continues the waste incinerator siting case study. We exploit a GIS for the evaluation of several decision criteria. A preference elicitation technique is then used to determine the preferred site. The technique, called Analytic Hierarchy Process (AHP) is one of the methods of multi-criterion decision analysis. It marries very well with a GIS and provides a very useful way of transforming data into decision, i.e. of producing information. Quoted from University of Cambridge, Dept. of engineering, Institute for manufacturing : AHP is especially suitable for complex decisions which involve the comparison of decision elements which are difficult to quantify. It is based on the assumption that when faced with a complex decision the natural human reaction is to cluster the decision elements according to their common characteristics. It involves building a hierarchy (Ranking) of decision elements and then making comparisons between each possible pair in each cluster (as a matrix). This gives a weighting for each element within a cluster (or level of the hierarchy) and also a consistency ratio (useful for checking the consistency of the data). The Analytical Hierarchy Process Model was designed by TL Saaty as a decision making aid.
What you should learn to do After completing the section-1 of this lecture you should be able to 1. Design a multi-criterion decision analysis scheme 2. Use GIS to evaluate the criteria 3. Implement the AHP/GIS approach in the Excel environment. Table of contents
Multi-criterion decision analysis A variety of MCDA techniques Exercise 1 Preference elicitation Quiz 1 The Analytic Hierarchy Process Quiz 2 Integration of the DSS with a GIS Exercise 2 Exercise 3 What have we learned ? Further readings
Multi-criterion decision analysis In the first lecture we have seen how the geographical data could be processed to provide a better support to the analysis of the decision to make. We have obtained an identification of 5 admissible sites, we have estimated the yearly volume of municipal waste to transport and we have computed the transportation cost from each commune to each possible site. These are just a few elements of the amount of relevant information that may be extracted from the geographical data stored in the GIS.
interpreted in maps or transformed and treated through models and then interpreted in maps and finally transformed into indicators and criteria values. MCDA techniques will transform the many indicators and criteria obtained through a GIS approach into a recommendation for a course of action that should be the "preferred one" for the DM concerned.
Figure 1 : The MCDA and the GIS Back to TOC A variety of MCDA techniques There are many techniques that have been developed to help DMs to decide under many criteria. Among the most celebrated techniques let us cite
MAUT (Multi-attribute utility theory) : a technique based on the paradigm of decision tree and risk analysis based on cardinal utility. For a coupling of MAUT with a GIS see Keisler et al. ELECTRE : ELECTRE was originally developed by B. Roy to incorporate the fuzzy (imprecise and uncertain) nature of decision making, by using thresholds of indifference and preference. The presentation of ELECTRE will be further developed in one lecture of topic 1.4 dedicated to MCDA.
Compromise Programming (CP) : It is used to identify solutions that are closest to the ideal solution as determined by some measure of distance. The solutions identified to be closest to the ideal solution are called compromise solutions and constitute the compromise set. The ideal solution is one which provides the extreme value for each of the criteria considered in the analysis. For a coupling of CP with a GIS see Tkach et al.
Back to TOC Exercise 1 Select a city that you know well. Consider the problem of siting a waste incinerator in the vicinity of this city. Explain the different studies that the concerned DM should order and the different criteria and/or indicators that should be computed, in order to evaluate the possible courses of action. Detail as much as you can the type of analysis that could be performed on a GIS Back to TOC Preference elicitation
Criteria evaluation
The data comes from a series of measurements concerning a variety of phenomena. These measurements enter into the definition of criteria which influence the preference of the decision maker. Typically, a decision maker involved in a problem of locating an obnoxious facility will use many different criteria to evaluate different options. For example, the criteria are organised in three groups related to economics, environmental impact and political acceptance, respectively.
Economic criteria: investment cost, transportation cost, maintenance cost; Environmental efficiency criteria: air pollutant emissions, increase in population exposure to immissions, population exposure to noise nuisance; Socio-political acceptance: coping with the NIMBY effect, respect of zoning, conformity with urban long range development planning.
In each group several criteria will be used to "measure" different components of the economics, the environmental impact or the political acceptance of the project under study.
Population exposure to nuisance: obtained through buffer analysis and application of a (long term) pollution dispersion model, or of a noise model,... NIMBY : identification of "neinsger" communes through analysis of past elections. Etc...
http://www.expertchoice.com/hierarchon/references/reflist.htm. In Europe, other multicriterion decision analysis tools have been successful, like e.g. Electre. The AHP method is based on two principles: 1. Build a hierarchy of criteria, in the form of graph where, on the left end you represent the decision to make and on the right end you represent the alternatives among which you have to decide the preferred one. 2. At each node of the hierarchy perform a weighting, summing to 1, which gives the relative preferences of the decision maker at this level of the hierarchy for the object that are directly linked to the node. This weighting is realised through a sequence of pairwise comparisons from which a consistent normalised set of weights is deduced. HIPRE http://www.hipre.hut.fi/ is an AHP tool, developed by the Systems Analysis Laboratory of the Helsinki University of technology, that is available on the web. We have implemented a very "primitive" version of AHP in an Excel file site_criteria.xls that you can download. The details of the method will be exposed in a forthcoming lecture (Lecture 1.4.1). Let us illustrate it on our siting problem.
Pairwise comparison
The complexity in preference elicitation comes from the fact that our mind has difficulties to compare together more than two things. For example, suppose one wants to compare three alternatives in terms of one criterion and give a degree of preference. The task is already complex.
Figure 2 : Comparing three alternatives We can easily make pairwise comparisons. For example we compare alternatives 1 and 2. We can take the analogy of a scale and give a ratio between 1 and 9 that will represent our level of preference for the most valuable of the two alternatives. A ratio equal to 1
means that the two alternatives are equivalent. A ratio 9 means that alternative 1 weighs 9 time more than alternative 1. Indeed if we were weighing alternative 2 in comparison with alternative 1 we should obtain a ratio 1/9, in this case.
Figure 3 : Comparing alternatives 1 and 2 We can repeat the pair wise comparison for another pair of alternatives
We thus obtain a matrix of ratios between each of the alternatives in a series of pairwise comparisons. For example we obtain the results shown in Table 1. For the criterion under consideration, alternative 1 weighs twice as much as alternative 3; alternative 2 weighs thrice as much as alternative 1; alternative 2 weighs 6 times as much as alternative 3. The problem consists now to infer, from this table the relative weights of the three alternatives, from the point of view of the criterion considered here.
Criterion
Alternative 1 Alternative 2 Alternative 3
Alternative 1 1 3 1/2
Alternative 3 2 6 1
Table 1 : matrix of pairwise comparisons We notice that, if we normalise the columns we obtain the following values
Criterion
Alternative 1 Alternative 2 Alternative 3
Table 2 : Normalising the columns We therefore get the following relative weights for the three alternatives
Criterion
Sums
Now we have obtained three weights which represent the relative importance of the three criteria in the preference, according to the criterion under consideration. You may easily check that the weights shown in Table 3 are fully compatible with the ratios of table 1. This shows that we have been perfectly consistent in our pairwise comparisons. The fact is that it is difficult to be totally consistent when we compare many pairs of alternatives. We see now how to deal with limited consistency. For example we could have obtained the following ratios
Criterion
Alternative 1 Alternative 2 Alternative 3
Alternative 1 1 3 1/2
Alternative 3 2 4 1
In that case we normalize each column and sum over the columns to get
Criterion
Alternative 1
Alternative 1 0.222
Alternative 2 0.211
Alternative 3 0.286
Alternative 2 Alternative 3
0.667 0.111
0.632 0.158
0.571 0.143
Table 4 : The normalised columns These normalised columns are not identical. Averaging over the columns we obtain the relative weights of the alternatives.
Criterion
Alternative 1 Alternative 2 Alternative 3 Table 5 : The relative weights
In the excel implementation of AHP that we use for this lecture we systematically use this avergaing technique to produce the relative weights at different levels of the hierarchy that we shall describe shortly.
The construction of the criteria hierarchy is the most important step in the AHP procedure. In complex problems one may have up to seven levels in the hierarchy, and up to 9 subcriteria for each group criterion.
Figure 1.1.4: The criteria hierarchy This completes the first phase in the AHP implementation. The weighing process The following table contains the results of the successive weightings, obtained at each level of the hierarchy through a systematic set of pairwise comparisons. The method for achieving this weighing process will be described in Lecture 1.4.1. For the moment it suffices to understand that these weights have been identified by the decider as a representation of his own preferences. If you download the Excel file site_criteria.xls you will obtain all the details of the implementation for obtaining these weights. You will see that the process is based on a series of pairwise comparisons of different elements from the point of view of each criterion which is located one level above in the hierarchy. From the complete set of pairwise comparisons one deduces a relative weight for the different elements eveluated through a given criterion. A global weight, called "score" is finally obtained for the alternatives, by combining the different weights calculated at different levels of the hierarchy. You shoud notice that the weights underthe heading " Invest" sum to 1 (up to some rounding error). Similarly for any other second level criterion, like e.g. " Nimby". So for this decision maker, when it comes to compare the different alternatives in terms of investment cost, then the resulting wheights for the five admissible sites are 0.29, 0.11, 0.05, 0.12, 0.45 for ZIMEYZA, BOIS-DE-BAY, VELODROME, LES RUPIERES, LES CHENEVIERS, respectively.
You will also notice that the two sub-criteria of the group economy have respective weights 0.67 and 0.33. Similarly, in the environment group, the three sub-criteria have respective weights 0.32, 0.54, 0.14. In the Politics group the two sub-criteria are weighed 0.67 and 0.33 respectively. Finally the three groups of criteria have respective weights 0.61 for economy, 0.29 for environment, 0.10 for politics.
0.61 economy 0.67 Invest ZIMEYZA BOIS-DE-BAY VELODROME LES RUPIERES LES CHENEVIERS 0.29 0.11 0.05 0.12 0.45 0.61 economy 0.33 Transp_cost 0.36 0.08 0.31 0.19 0.04 0.29 environment 0.32 NO2_expo 0.41 0.26 0.04 0.08 0.21 0.29 environment 0.54 Transp_nuis 0.35 0.08 0.30 0.18 0.04 0.29 environment 0.14 Bulk_transp 0.21 0.05 0.21 0.05 0.47 0.10 Politics 0.67 Nimby 0.29 0.11 0.05 0.12 0.45 0.10 Politics 0.33 Neighborhood 0.36 0.08 0.31 0.19 0.04
Table 1.1.1: The weights at each level of the hierarchy Now, to obtain the scores of the different alternatives in the preference of the decision maker, one combines these different weights together. This will maintain the normalisation (summing to 1). The result is shown below, first in Table 1.1.2 where the composite weights are detailed at the level of the sub-criteria and then in in Table 1.1.3 where the composite scores are shown for the different alternatives.
0.05 0.18
0.04 0.01
0.01 0.02
0.03 0.01
0.00 0.02
0.01 0.03
0.01 0.00
Table 1.1.2: The composite weights Finally, from these composite weights one obtains, through summation, the global scores of the 5 different possible sites.
Global scores ZIMEYZA BOIS-DE-BAY VELODROME LES RUPIERES LES CHENEVIERS 0.32 0.11 0.15 0.14 0.27
Table 1.1.3: The global scores The result show that the location ZIMEYZA is ranked first (score 0.32), followed by LES CHENEVIERS (score 0.27), etc.... Combining GIS operations to evaluate each criterion with an MCDA technique we have arrived at a ranking of the 5 possible alternatives. The GIS approach has been integrated in a decision support system. Back to TOC Quiz 2 The AHP method is particularly helpful because 1. It gives weights that are normalised to sum to 1. 2. It permits to combine together qualitative and quantitative criteria through a simple sequence of pairwise comparisons. 3. It corresponds to the psychology of most decision makers. A:2 Back to TOC
Integration of the DSS with a GIS The data treatment to extract valuable information concerning the waste incinerator siting problem has been done with ArcView. The AHP implementation has been performed with the help of Excel. There are possibilities to connect a GIS and a DSS in a common environment. We show below a map that has been constructed in the Excel environment and which represents the result of the AHP implementation. The size of each pie is proportional to the score of the location; the different components corresponding to each sub-criterion are shown in different colours. This summarises nicely the analysis.
Figure 1.1.5: The map showing the scores of the different sites Exercise 2 Use the excel file site_criteria.xls to define your own preferences regarding the proposed criteria. Compare the weights and the scores obtained with the ones indicated above. Back to TOC Exercise 3
Download the hands-on exercise Hands_On_Lect1_1_2.doc Thematic Maps in Excel. What have we learned? We have implemented a multi-criterion decision analysis technique for gaining information concerning the decision of where to locate a waste incinerator. An implementation of the AHP method for multi-criterion decision analysis has shown the possible integration of a GIS and a DSS, where the data is prepared in a GIS, an analysis is performed with the help of a systematic evaluation method and the result of the analysis is displayed in the GIS again.
Further readings
Keisler J. M. and Sundell R. C., Combining Multi-Attribute Utility and Geographic Information for Boundary Decisions: An Application to Park Planning Journal of Geographic Information and Decision Analysis Volume 1, No 2, 1997. Back to citation point Roy B, The outranking approach and the foundation of ELECTRE methods, Theory and Decision, Vol. 31, pp. 49-73, 1991. Back to citation point Tkach R.J. and Simonovic S.P., A New Approach to Multi-criteria Decision Making in Water Resources, Journal of Geographic Information and Decision Analysis, Vol.1, no.1, pp. 25-43, 1997. Back to citation point Saaty TL, The Analytic Hierarchy Process, NY, McGraw Hill,1980. Back to citation point
Back to TOC