Upload
emma-lane
View
223
Download
0
Tags:
Embed Size (px)
Citation preview
Robustness in assessment of strategic transport projectsThe 21st International Conference on Multiple Criteria Decision MakingJyväskyläJune 13-17. 2011
Anders Vestergaard JensenDepartment of TransportTechnical University of Denmark
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 20112 15.06.2011
Introduction• The paper is developed as part of the Ph.D study: Appraisal of Transport
Projects: Assessing Robustness in Decision Making
• The focus is on:– Dealing with subjectivity in assessment of larger transport projects– Transparent decision making process– Consensus– Communicative issues
• The background can be seen from the following statements– Policy-making has entered an era in which societal benefits of
governmental actions are increasingly questioned.
– Multi criteria analysis (MCA) is of growing importance due to inclusion of non-economic factors, MCA criteria weights are in this respect the decisive factors in decision-making.
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 20113
Strategic decision making• Recognizing that formal economic evaluation analyses tend to be
inadequate – because objectives often are broader than pure economic or market concerns
• Focus is on transparent decision making processes
• For public decision making - decisions need to be justifiable
• There is a need to research the subjective part of the decision processes and the role of the decision support
15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 20114
Appraisal approach
15.06.2011
• Assumption: Finite set of alternatives Idealized preference structure (participants are assumed to avoid
strategic behavior)
• Methodology Pair wise comparisons Multiplicative AHP (REMBRANDT) Rank order weights (ROD)
• Decision conference Stakeholder preferences Tailored to specific decision problem
• Robustness analysis• Weight stability interval
Group processes Technical
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 20115
Framework
15.06.2011
Decision Conference
Definition of MCA Criteria
Pair wise comparisons using
REMBRANDT
Weighting of criteria using rank
order
Set of alternatives
Cost Benefit Analysis
Multi Criteria Analysis
Robustness Analysis
Input to the decision making process
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 20116
Robustness Analysis• Robustness of the assessment of alternatives is in this context explored
by focusing on the uncertainties in the different model components
• Specifically the uncertainty relating to the criteria weights is examined
• This examination is carried out by modeling the influence that the weights have on the preference order of the alternatives
• The purpose is to create understanding and consensus of the decision making outcome
15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 20117
Weight stability intervals (WSI)Investigate how the desirability of the alternatives changes when all the
weights of the criteria are kept constant except for one criterion.
We change 1 criterion (Ci):
All other weights are adjusted so only the importance of Ci relative to the other criteria is modified:
and are related and constrained as follows:
and
15.06.2011
ii ww )1('
jj ww '
i
i
w
w
1
)1(1
i
i
w
w
11
iw
1
10
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 20118
Case: Elsinore-Helsingborg Connection• Alternatives (all tunnels):
15.06.2011
Alternatives
A1 Passenger train – Public transport
A2 Passenger train and car (dual tunnel)
A3 Passenger and goods train. car (dual tunnel)
A4 Passenger and goods train. car (single tunnel)
All alternatives have been found to be socio-economic feasible
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 20119
Appraisal• Criteria:
– Construction cost, maintenance, scrap value, time savings, operating
costs, emissions, revenue (tickets)
– Impact on towns and land-use
– Regional economy
– Transport network and accessibility
– Impact on greening of transport
– Impact on flexibility in logistics
15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201110
Pair wise comparison
Criterion: Socio-economic robustness
Alt1 Alt2 Alt3 Alt4 GeoMean
Alt1 0 -6 -5 -7 0.044
Alt2 6 0 1 -1 2.828
Alt3 5 -1 0 -2 1.414
Alt4 7 1 2 0 5.657
15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201111
Overall score
Criteria Criterion weight
The alternative score within criterion
Alt1 Alt2 Alt3 Alt4
Socio-economic robustness 0.297 0.044 2.828 1.414 5.657
Town and land-use 0.139 0.022 1.414 5.657 5.657
Regional economics 0.188 0.074 0.841 6.727 2.378
Transport network and accessibility
0.241 0.074 0.841 6.727 2.378
Greening of transport 0.091 0.595 0.037 8.000 5.657
Flexibility on logistics 0.045 0.022 0.707 8.000 8.000
Weighted score: 0.062 0.968 4.233 3.960
15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201112
Weights assessed by the DMs
15.06.2011
Socio-economic robustness
Town and land-use
Regional economics
Transport network and accessibility
Greening of transport
Flexibility on logistics
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
GroupEDCBA
Criteria Weights
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201113
Weight intervals
15.06.2011
Socio
-eco
nom
ic ro
bustne
ss
Town
and
land
-use
Regiona
l eco
nom
ics
Tran
spor
t net
work an
d ac
cess
ibilit
y
Green
ing
of tr
ansp
ort
Flexib
ility on
logistics
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Stated weight intervals(min. and max. of all individual weights)
We
igh
t
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201114
Weight stability
15.06.2011
0
0.05
932
0.11
864
0.17
796
0.23
728
0.29
66
0.35
592
0.41
524
0.47
456
0.53
388
0.59
32
0.65
252
0.71
184
0.77
116
0.83
048
0.88
98
0.94
912
0
2
4
6
8
Criterion: Socio-economic robustness
Alt1Alt2Alt3Alt4
Tota
l score
Weight assessed by the group
Upp
er
and
low
er
weig
hts
Weight
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201115
Weight Stability Interval• In the interval defined by the lower and upper the preference order does
not change
15.06.2011
Criteria Group weight
Lower stable
Upper stable
Socio-economic robustness 0.297 0 0.329
Town and land-use 0.139 0 1
Regional economics 0.188 0.132 1
Transport network and accessibility 0.241 0.188 1
Greening of transport 0.091 0 0.545
Flexibility on logistics 0.045 0 1
Top ranked alternatives
shift preference
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201116 15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201117 15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201118 15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201119 15.06.2011
Socio
-eco
nom
ic ro
bustne
ss
Town
and
land
-use
Regiona
l eco
nom
ics
Tran
spor
t net
work an
d ac
cess
ibilit
y
Green
ing
of tr
ansp
ort
Flexib
ility on
logistics
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Stated weight intervals(min. and max. of all individual weights)
We
igh
t
Weight stability intervals
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201120
Framework
15.06.2011
Decision Conference
Definition of MCA Criteria
Pair wise comparisons using
REMBRANDT
Weighting of criteria using rank
order
Set of alternatives
Cost Benefit Analysis
Multi Criteria Analysis
Robustness Analysis
Input to the decision making process
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201121
Conclusions (I)Case:• A small increase on the weight on the CBA will result in a change of
preference order– Within the stated weight range from the DMs
• A minor decrease in the weights for Regional economic and Transport network and accessibility will change the preference order
– Within the stated weight range from the DMs
• Variations in 3 out of 6 criteria weights will not alter the preference order of the top ranked alternatives
• Only changes in 2 criteria weights can influence on the top preferred alternatives
15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201122
Conclusions (II)Methodology:• Possible to determine if changes in the criteria weights are critical for the
preference order of alternatives
• Transparent analysis, which can be communicated to DMs
• Assessing the impact of the subjective weights
Perspectives:• A ‘full scale’ decision conference is to be held with experts in October
• Modification of methodology when tested on ‘real’ decision makers
15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201123
The work presented in these slides is part of the Øresund EcoMobility project and co-funded by the EU / Interreg IV A ØKS Programme
http://www.oresundecomobility.org15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201124 15.06.2011
Thank you…
15.06.2011
DTU Transport. Technical University of Denmark Anders Vestergaard Jensen - MCDM 201126
Socio-economic robustness criterion
15.06.2011
Alternative B/C rate Socio-economic robustness
Alt1 1.23 27%
Alt2 2.69 77%
Alt3 2.25 69%
Alt4 3.10 82%
DTU Transport. Technical University of Denmark MCDM 2011, Jyväskylä27
REMBRANDT calculations
• A1, A2, A3 and A4, are compared in a pair wise way under the five criteria
• The final scores are calculated using the multiplicative model and normalised:A1: 0.710.19 . 4.000.06 . 1.000.27 . 4.000.12 . 0.500.36 = 0.93 ~ 0.12A2: 0.040.19 . 4.000.06 . 0.030.27 . 16.000.12 . 0.030.36 = 0.09 ~ 0.01A3: 11.310.19 . 0.250.06 . 16.000.27 . 0.060.12 . 4.000.36 = 3.69 ~ 0.47A4: 2.830.19 . 0.250.06 . 2.000.27 . 0.250.12 . 16.000.36 = 3.12 ~ 0.40
13-06-2011
Pair wise comparisons (δjk) Transformations (γ = 0.7)Geo.mean
A1 A2 A3 A4 A1 A2 A3 A4 ScoreA1 0 4 -4 -2 1 16 0.0625 0.25 0.71A2 -4 0 -8 -6 0.0625 1 0.0039 0.0156 0.04A3 4 8 0 2 16 256 1 4 11.31A4 2 6 -2 0 4 64 0.25 1 2.83