partitioning of public transit networks - [bachelor's...
TRANSCRIPT
![Page 1: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/1.jpg)
Partitioning of Public Transit Networks[Bachelor’s thesis]
Matthias Hertel
Albert-Ludwigs-Universitat Freiburg
11.09.2015
![Page 2: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/2.jpg)
Introduction Data model Algorithms Evaluation Appendix
Contents
1 IntroductionMotivationGoal
2 Data model
3 AlgorithmsK-meansMerging algorithmMETISPUNCH
4 Evaluation
![Page 3: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/3.jpg)
Introduction Data model Algorithms Evaluation Appendix
Transfer Patterns [1] with partitioning
Transfer Patterns = sequences of transfers on optimal routesFreiburg → Zurich: {[Freiburg, Zurich], [Freiburg, Basel, Zurich]}
Compute Transfer Patterns between
stations of the same partition
border stations b(Cx ) and b(Cy )
⇒ reduced runtime⇒ reduced space
Query “A → B”:A → b(CA)→ b(CB)→ B⇒ little slower query times
![Page 4: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/4.jpg)
Introduction Data model Algorithms Evaluation Appendix
Transfer Patterns [1] with partitioning
Transfer Patterns = sequences of transfers on optimal routesFreiburg → Zurich: {[Freiburg, Zurich], [Freiburg, Basel, Zurich]}
Compute Transfer Patterns between
stations of the same partition
border stations b(Cx ) and b(Cy )
⇒ reduced runtime⇒ reduced space
Query “A → B”:A → b(CA)→ b(CB)→ B⇒ little slower query times
![Page 5: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/5.jpg)
Introduction Data model Algorithms Evaluation Appendix
Transfer Patterns [1] with partitioning
Transfer Patterns = sequences of transfers on optimal routesFreiburg → Zurich: {[Freiburg, Zurich], [Freiburg, Basel, Zurich]}
Compute Transfer Patterns between
stations of the same partition
border stations b(Cx ) and b(Cy )
⇒ reduced runtime⇒ reduced space
Query “A → B”:A → b(CA)→ b(CB)→ B⇒ little slower query times
![Page 6: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/6.jpg)
Introduction Data model Algorithms Evaluation Appendix
Goal
Partition the stations of a public transit network, such that
partitions are small
most traffic lies inside the partitions
![Page 7: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/7.jpg)
Introduction Data model Algorithms Evaluation Appendix
Dataset
schedule of Deutsche Bahn (2015)
only local traffic (no ICEs and ICs)
modelled as undirected weighted graph
stations → nodes
connections → edges
frequencies → edge weights
heuristical footpaths (distance ≤ 400 m; weight 200,000)
![Page 8: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/8.jpg)
Introduction Data model Algorithms Evaluation Appendix
K-means-clustering [2]
uses only geographic data
Algorithm 1 k-means-clustering
initializewhile assignments change do
update assignmentsupdate means
end while
![Page 9: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/9.jpg)
Introduction Data model Algorithms Evaluation Appendix
Merging algorithm [3]
hierarchical
merges neighboured partitions
hyperparameter k = number of partitions
hyperparameter U = upper bound partition size
order distinguished by a utility function
f (u, v) = 1s(u)·s(v) · (
w(u,v)√s(u)
+ w(u,v)√s(v)
)
s(u) = size of us(v) = size of vw(u, v) = sum of edge weights between u and v
![Page 10: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/10.jpg)
Introduction Data model Algorithms Evaluation Appendix
Merging algorithm [3]
hierarchical
merges neighboured partitions
hyperparameter k = number of partitions
hyperparameter U = upper bound partition size
order distinguished by a utility function
f (u, v) = 1s(u)·s(v) · (
w(u,v)√s(u)
+ w(u,v)√s(v)
)
s(u) = size of us(v) = size of vw(u, v) = sum of edge weights between u and v
![Page 11: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/11.jpg)
Introduction Data model Algorithms Evaluation Appendix
METIS [4]
graph partitioning framework
state of the art
can be downloaded 1
hyperparameter k = number of partitions
three phases (next slide)
1http://glaros.dtc.umn.edu/gkhome/metis/metis/download
![Page 12: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/12.jpg)
Introduction Data model Algorithms Evaluation Appendix
METIS
Figure : The three phases of METIS (Source: [4])
![Page 13: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/13.jpg)
Introduction Data model Algorithms Evaluation Appendix
PUNCH [5]
“partitioning using natural cut heuristics”
hyperparameter U = upper bound partition size
two phases
filtering phaseassembly phase
![Page 14: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/14.jpg)
Introduction Data model Algorithms Evaluation Appendix
PUNCH
Filtering phase: contract regions that are separated by small cuts
main graph
1
< U
bridge⇒
main graph
≤ U
![Page 15: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/15.jpg)
Introduction Data model Algorithms Evaluation Appendix
PUNCH
Assembly phase
initial solution: run merging algorithm on filtered graph
local optimization:
uncontract small regionsrerun merging algorithmtake better solution
![Page 16: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/16.jpg)
Introduction Data model Algorithms Evaluation Appendix
Comparison: cut size
maximum partition size
0 5000 10000 15000 20000 25000 30000 35000
cut siz
e
10 6
10 7
10 8
10 9
10 10
k-means
merging
PUNCH
METIS
Figure : Cut size over maximum partition size.
![Page 17: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/17.jpg)
Introduction Data model Algorithms Evaluation Appendix
Comparison: cut edges
maximum partition size
0 5000 10000 15000 20000 25000 30000 35000
num
ber
of cut edges
1000
10000
100000
k-means
merging
PUNCH
METIS
Figure : Cut edges over maximum partition size.
![Page 18: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/18.jpg)
Introduction Data model Algorithms Evaluation Appendix
PUNCH - unweighted graph
main graph
1
1 1
10
1
11
11
⇒
main graph
13
1
⇒ minimum cut preserved
![Page 19: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/19.jpg)
Introduction Data model Algorithms Evaluation Appendix
PUNCH - unweighted graph
main graph
1
1 1
10
1
11
11
⇒
main graph
13
1
⇒ minimum cut preserved
![Page 20: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/20.jpg)
Introduction Data model Algorithms Evaluation Appendix
PUNCH - weighted graph
main graph
1
1 1
10
200,000
11
11
⇒
main graph
13
200,000
⇒ minimum cut not preserved
![Page 21: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/21.jpg)
Introduction Data model Algorithms Evaluation Appendix
PUNCH - weighted graph
main graph
1
1 1
10
200,000
11
11
⇒
main graph
13
200,000
⇒ minimum cut not preserved
![Page 22: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/22.jpg)
Introduction Data model Algorithms Evaluation Appendix
The gain of footpaths
merging algorithm with U=4,000
Figure : no footpaths
Figure : with footpaths
![Page 23: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/23.jpg)
Introduction Data model Algorithms Evaluation Appendix
The gain of footpaths
merging algorithm with U=4,000
Figure : no footpaths Figure : with footpaths
![Page 24: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/24.jpg)
Introduction Data model Algorithms Evaluation Appendix
Conclusions
K-means better than expected ⇒ traffic geographicallyclustered
merging algorithm and METIS produce good results
arbitrary utility functions can be used with the mergingalgorithm
PUNCH: filtering phase must use edge weights
footpaths prohibit geographically overlapping partitions
![Page 25: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/25.jpg)
Introduction Data model Algorithms Evaluation Appendix
Conclusions
K-means better than expected ⇒ traffic geographicallyclustered
merging algorithm and METIS produce good results
arbitrary utility functions can be used with the mergingalgorithm
PUNCH: filtering phase must use edge weights
footpaths prohibit geographically overlapping partitions
![Page 26: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/26.jpg)
Introduction Data model Algorithms Evaluation Appendix
Conclusions
K-means better than expected ⇒ traffic geographicallyclustered
merging algorithm and METIS produce good results
arbitrary utility functions can be used with the mergingalgorithm
PUNCH: filtering phase must use edge weights
footpaths prohibit geographically overlapping partitions
![Page 27: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/27.jpg)
Introduction Data model Algorithms Evaluation Appendix
Conclusions
K-means better than expected ⇒ traffic geographicallyclustered
merging algorithm and METIS produce good results
arbitrary utility functions can be used with the mergingalgorithm
PUNCH: filtering phase must use edge weights
footpaths prohibit geographically overlapping partitions
![Page 28: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/28.jpg)
Introduction Data model Algorithms Evaluation Appendix
Conclusions
K-means better than expected ⇒ traffic geographicallyclustered
merging algorithm and METIS produce good results
arbitrary utility functions can be used with the mergingalgorithm
PUNCH: filtering phase must use edge weights
footpaths prohibit geographically overlapping partitions
![Page 29: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/29.jpg)
Introduction Data model Algorithms Evaluation Appendix
Questions?
Thank you for your attention!
![Page 30: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/30.jpg)
Introduction Data model Algorithms Evaluation Appendix
Bibliography
H. Bast, E. Carlsson, A. Eigenwillig, R. Geisberger, C. Harrelson, V. Raychev, and F. Viger, “Fast routing in
very large public transportation networks using transfer patterns,” in Algorithms–ESA 2010. Springer,2010, pp. 290–301.
J. MacQueen, “Some methods for classification and analysis of multivariate observations,” in Proceedings
of the fifth Berkeley symposium on mathematical statistics and probability, vol. 1, no. 14. Oakland, CA,USA., 1967, pp. 281–297.
M. G. van der Horst, “Optimal route planning for car navigation systems,” Master’s thesis, Technische
Universitaet Eindhoven, 2003.
G. Karypis and V. Kumar, “A fast and high quality multilevel scheme for partitioning irregular graphs,”
SIAM Journal on scientific Computing, vol. 20, no. 1, pp. 359–392, 1998.
D. Delling, A. V. Goldberg, I. Razenshteyn, and R. F. Werneck, “Graph partitioning with natural cuts,” in
Parallel & Distributed Processing Symposium (IPDPS), 2011 IEEE International. IEEE, 2011, pp.1135–1146.
![Page 31: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/31.jpg)
Introduction Data model Algorithms Evaluation Appendix
Cut size with k-means
round
0 50 100 150 200 250
cut
siz
e
×108
1
2
3
4
5
6
7
8
9
10
11
k=100
k=200
k=400
k=800
Figure : Cut size over maximum partition size.
![Page 32: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/32.jpg)
Introduction Data model Algorithms Evaluation Appendix
METIS
unbalancing ratio r
s(p) ≤ r · Nk
number of partitions
10 100 1000
ma
xim
um
pa
rtitio
n s
ize
0
5000
10000
15000
20000
25000
30000
metis, r=1.5
metis, r=2.0
metis, r=4.0
Figure : Maximum partition size over number of partitions.
![Page 33: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/33.jpg)
Introduction Data model Algorithms Evaluation Appendix
METIS
number of partitions
10 100 1000
cut siz
e
×10 7
0
2
4
6
8
10
12
14
16
18
metis, r=1.5
metis, r=2.0
metis, r=4.0
Figure : Cut size over number of partitions.
![Page 34: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/34.jpg)
Introduction Data model Algorithms Evaluation Appendix
PUNCH
Filtering phase, pass 1: contract bridge-separated regions
main graph
1
1 1
10
bridge
⇒
main graph
13
![Page 35: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/35.jpg)
Introduction Data model Algorithms Evaluation Appendix
PUNCH
Filtering phase, pass 2: contract simple paths
main graph
1
1
1
1
1
⇒main graph
5
![Page 36: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/36.jpg)
Introduction Data model Algorithms Evaluation Appendix
PUNCH
Filtering phase, pass 3: contract two-cut-separated regions
main graph
1 1
10
⇒
main graph
12
![Page 37: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/37.jpg)
Introduction Data model Algorithms Evaluation Appendix
PUNCH
Filtering phase, pass 4: contract “natural cut”-separated regions
Figure : Finding a “natural cut” (Source: [5])
![Page 38: Partitioning of Public Transit Networks - [Bachelor's thesis]ad-publications.informatik.uni-freiburg.de/.../Bachelor...presentation.p… · [Bachelor’s thesis] Matthias Hertel Albert-Ludwigs-Universit](https://reader033.vdocuments.site/reader033/viewer/2022042909/5f3b3c0d9a4b7716a36b2752/html5/thumbnails/38.jpg)
Introduction Data model Algorithms Evaluation Appendix
k-means merging PUNCH METIS
partitions 181 181 176 181max. part. size 4,015 1,873 1,975 3,132cut size 154.7·106 42.8·106 496.4·106 45.5·106
cut edges 12,273 9,497 13,917 8,562cut edges (%) 2.2 1.7 2.5 1.6border nodes 15,564 12,954 17,669 12,010border nodes (%) 6.2 5.2 7.1 4.8runtime (s) 53.8 2.9 118.3 0.3
Table : Results of the four algorithms with about 181 partitions.