Você está na página 1de 13

6/2/2014

1
ABMApplications
Ut S j P t Utomo Sarjono Putro
ABMLecture
Analyticalvs Simulation
2
MathematicalModels
3
Decision Support Systems and Intelligent Systems, EfraimTurbanandJ ayE. Aronson, 6thedition,
Copyright 2001, PrenticeHall, Upper SaddleRiver, NJ
4
6/2/2014
2
TravelingSalesmanProblem
TheTravelingSalesmanProblem isdefinedas:
Wearegivenasetofcitiesandasymmetricdistancematrix
thatindicatesthecostoftravelfromeachcitytoeveryother
city.
Thegoalistofindtheshortestcirculartour,visitingeverycity
exactlyonce,soastominimizethetotaltravelcost,which
i l d th t f t li f th l t it b k t th fi t
TravelingSalesmanProblem 5
includesthecostoftravelingfromthelastcitybacktothefirst
city.
Distance/CostMatrixForTSP
1 2 3 4 5 6
1 0 863 1987 1407 998 1369
2 863 0 1124 1012 1049 1083
3 1987 1124 0 1461 1881 1676
4 1407 1012 1461 0 2061 2095
5 998 1049 1881 2061 0 331
6
Costmatrixforsixcityexample.
DistancesinKilometers
5 998 1049 1881 2061 0 331
6 1369 1083 1676 2095 331 0
TSPwith6cities
Irepresenteverycitywithaninteger.
C id 6 I di i i Consider6Indiancities
Mumbai,Nagpur,Calcutta,Delhi,BangaloreandChennai
andassignanumbertoeach.
Mumbai1
Nagpur 2
7
Calcutta 3
Delhi4
Bangalore5
Chennai6
Alternativesofroutes
Thusapathwouldberepresentedasasequence ofintegers
from 1 to 6. from1to6.
Thepath[123456] representsapathfromMumbaito
Nagpur,NagpurtoCalcutta,CalcuttatoDelhi,Delhito
Bangalore,BangaloretoChennai,andfinallyfromChennaito
Mumbai.
ThisisanexampleofPermutationEncoding asthepositionof
the elements determines the fitness of the solution
8
theelementsdeterminesthefitnessofthesolution.
Thisisknowntobeahardproblemtosolvebecausethe
numberofpossibleroutesisN!whereN=thenumberof
cities
6/2/2014
3
AnalyticalSolutiontoTSP
AnalyticalsolutionencodedinVisualC++onaPentium4personal
computerwithspeed3GHzand448MBRAMunderMSWindowsXP.
Graphic plotted on logarithmic scale Graphicplottedonlogarithmicscale
1.00E02
1.00E+00
1.00E+02
1.00E+04
1.00E+06
1.00E+08
0 5 10 15 20 25
im
e
in
S
e
c
o
n
s
Timetofindoptimum
solutionfor20citiesis
morethan19years
1.00E10
1.00E08
1.00E06
1.00E04
T
Number ofCities
GraphicisadaptedfromZakir H.Ahmed,GeneticAlgorithmfortheTravelingSalesmanProblemusingSequential
ConstructiveCrossoverOperator,InternationalJournalofBiometrics&Bioinformatics(IJBB)Volume(3):Issue(6)
UsingGAbasedSimulationforTSP
Current
Population(t)
Next
Population
(t+1)
Survivalofthefittest
NaturalSelection
Encoding,Selection,Reproduction
Thefittest
Population
(endof
simulation)
Encoding
Irepresenteverycitywithaninteger.
C id 6 I di i i Consider6Indiancities
Mumbai,Nagpur,Calcutta,Delhi,BangaloreandChennai
andassignanumbertoeach.
Mumbai1
Nagpur 2
11
Calcutta 3
Delhi4
Bangalore5
Chennai6
Encoding(contd.)
Thusapathwouldberepresentedasasequence ofintegers
from 1 to 6. from1to6.
Thepath[123456] representsapathfromMumbaito
Nagpur,NagpurtoCalcutta,CalcuttatoDelhi,Delhito
Bangalore,BangaloretoChennai,andfinallyfromChennaito
Mumbai.
12
6/2/2014
4
FitnessFunction
Thefitnessfunctionwillbethetotalcostofthetour
t d b h h representedbyeachchromosome.
Thiscanbecalculatedasthesumofthedistances traversedin
eachtravelsegment.
The Lesser The Sum The Fitter The Solution
13
TheLesserTheSum,TheFitterTheSolution
RepresentedByThatChromosome.
Distance/CostMatrixForTSP
1 2 3 4 5 6
1 0 863 1987 1407 998 1369
2 863 0 1124 1012 1049 1083
3 1987 1124 0 1461 1881 1676
4 1407 1012 1461 0 2061 2095
5 998 1049 1881 2061 0 331
14
Costmatrixforsixcityexample.
DistancesinKilometers
5 998 1049 1881 2061 0 331
6 1369 1083 1676 2095 331 0
FitnessFunction (contd.)
So,forachromosome[413256],thetotalcostoftravelor
fitnesswillbecalculatedasshownbelow
Fitness =1407+1987+1124+1049+331+2095
=7993kms.
SinceourobjectiveistoMinimize thedistance,thelesserthe
l d h f h l
15
totaldistance,thefitterthesolution.
Mainidea: "SelectTheBest,DiscardTheRest.
Reproduction
Theprocessthatchoosessolutionstobepreserved
andallowedtoreproduceandselectswhichones
musttodieout.
Themaingoal ofthereproductionoperatoristoemphasize
16
Themaingoal ofthereproductionoperatoristoemphasize
thegoodsolutions andeliminatethebadsolutions ina
population(whilekeepingthepopulationsizeconstant )
6/2/2014
5
So,howtoselectthebest?
RouletteSelection
RankSelection
17
TournamentSelection
Main idea: the fitter is the solution with the most
Roulettewheelselection
Mainidea:thefitteristhesolutionwiththemost
chancestobechosen
HOW IT WORKS ?
18
HOWITWORKS?
No. String Fitness % Of Total
ExampleofRoulettewheelselection
1 01101 169 14.4
2 11000 576 49.2
3 01000 64 5.5
19
4
10011 361 30.9
Total
1170 100.0
Citation: :www.cs.vu.nl/~gusz/
Roulettewheelselection
Chromosome1
Chromosome 2
Chromosome 3
Chromosome 4
20
Allyouhavetodoisspintheballandgrabthechromosomeatthe
pointitstops
6/2/2014
6
Mainidea:combinegeneticmaterial(bits)of2
parentchromosomes(solutions)and
d hild i h t i ti
Crossover
produceanewchildpossessingcharacteristics
ofbothparents.
Howitworks?
21
Severalmethods.
SinglePointCrossover Arandompointischosenonthe
individualchromosomes(strings)andthegenetic
materialisexchangedatthispoint.
Crossovermethods
22
Mainidea: randominversionofbitsinsolution
tomaintaindiversityinpopulationset
Mutation
y p p
23
Ex.giraffes mutationscouldbebeneficial.
Citation: http://www.ewh.ieee.org/soc/es/May2001/14/MUTATE0.GIF
SpectrumofAgentBasedModel
Positive
Developandstrengthen
theoreticalunderstanding
Testingatheory
Strengthen mental model
Normative
Policydesign
Analysisconsequencesof
policyimplementation
Scenario planning
Strengthenmentalmodel
andintuition
Scenarioplanning
6/2/2014
7
ExamplesofABMmodelling and
simulation
Security
PatrolMaritime
AntiBio
Terrorism
Social
Pandemic
Prevention
Education
Energy
Electricity
Efficiency
BBM
Distribution
Agricultural
Transportation
Traffic
EvacuationRoute
AntiBirdFluPandemicSimulation
Deskripsi:
ThismodelistodeveloppolicytopreventbirdflupandemicinBandung
Spektrum: Normative
Input:
Datademography
Datacitystructure
Datapeopleactivity
DataAviandistribution
Spektrum: Normative
Tahap Pengembagan: Advance
Pengembang : TIT;SBMITB
Output:
Estimationofnumberof
peoplewhoareinfected,stay
inhospital,anddeadforeach
policy.
SystemsResearchandBehavioralScience,Vol.25,No.5
TransitionmodelofBirdFluVirus
Dead
4
3days
Dead 1
Recovery
1
Infection
2 0,8
2m
3days 2days
3s 0,2
3
0,8
3days
0,2
4m
0,8
1
3days
4c
0,2
Dead 1
Recovered
27
3m
0
2m
2days
2days
1
3s
1day
5
1
1
2days
1
BandungPeopleActivityModel
Policies ??
28
6/2/2014
8
ResultsofSimulation
Policies:
Closingschool
Usingmask
Disinfectantusing
Pengaturan kepadatan ruang
Strategi vaksinasi
InfectedPeople
DeadPeople
Output:
Numberofpeoplewhoare
infectedanddeadforeachpolicy
combination.
29
AntiSmallpoxBioTerrorismSimulation
Deskripsi:
ThismodelModel istoformulatepolicytominimizenegativeimpactsofsmallpoxbio
terrorism.
Spektrum Normative
Input:
Datademography
Datacitystructure
Datapeopleactivities
Spektrum: Normative
Developmentstage: Advance
Creator: TIT;SBMITB
Output:
Numberofpeoplewhoare
infectedanddeadforeach
policy.
Simulationofrecruitingsystemfor
newhighschoolstudents
Description:
Thismodelistoformulaterecruitmentsystemfornewhighschoolstudentsinorderto
minimizeschoolsthatfinallyhaslackofstudents
Spectrum: Normative
Input:
Datapeopledistribution
Dataschooldisribution
Datacapacityandhistoryof
studentsacceptedinhigh
schools in Bandung
Spectrum: Normative
Developmentstage: Advance
Creator: SBMITB
0.100
0.150
0.200
0.250
0.300
0.350
0.400
Enrollmentdistributionineachcluster
Output:
Estimationofstudent
distributionineachschoolfor
eachscenariosofcluster
numbers,numberofschools
canbeelected
schoolsinBandung.
0.000
0.050
0.100
1 2 3 4
Equal Neutral Pessimistic Optimistic
Simulationofhouseholdelectricity
consumptionofinBandung
Description:
Thismodelistoformulatepolicyaboutpricetariffforeachhouseholdconsumers
segmentsinordertopromoteelectricityefficiency.
Input:
Datasegmentelectricty segment
Datamonthlyloadforeach
segment
Data household electricuty
Spektrum: Normative
Developmentstage: Medium
Creator: SBMITB
1500000
2000000
2500000
MonthlyElectricityConsumption(kWH)
+10%TariffScenario
R1450
R1900
R11300
Output:
Estimationofloadusedper
scenarioTDL.
Datahouseholdelectricuty
appliances(survey)
0
500000
1000000
0 10 20 30 40 50 60 70
R12200
R2
R3
6/2/2014
9
SimulationofBBMDistributionin
Indonesia
Description:
ThismodelistostudybehaviorpatternthatinfluenceBBMsmugglingtoforeign
countriesandeffortsnecessarytominimizeit.
Input:
DatadistributionchannelBBM
fromPERTAMINA
Spectrum: Positive
Developmentstage: Advance
Creator: SBMITB
Output:
EstimationofnumberofBBM
smuggledtoforeigncountries
foreachbehaviorpattern.
Evacuationsimulationforthecityof
Padanginthecaseoftsunami
Description:
Thismodelistodesignevacuationscenariotominimizenumberofvictims.Evacuation
processisstartedin5minutesafteradetectedtsunami.
Input:
Datageography
Datanumberanddistributionofpeople.
Evacuationscenario
Spectrum: Positive
Developmentstage: initialstage
Creator: BerlinInstituteof
Technology
Output:
Estimationofnumberofevacuated
people,numberofvictims,outflow
rateforeachleveloftsunamiand
evacuationroute.
TrafficSimulation
Description:
ThismodelistodesigntrafficsysteminBerlinCitytodecreasetrafficjam.
Input:
Dataroadstructure
Datapeopledistribution
Spectrum: Positive
Developmentstage:initialstage
Creator: BerlinInstituteof
Technology,ETHZurich
Output:
Estimationofvehiclespeedinsegmentof
roadsforeachtrafficsettings.
Datadailyrouteofpeople(traveldemand)
fromsurvey
AgentBasedSimulationofAgricultural
Policies
Description:
Thismodelistoanalyzeimpactofagriculturalpolicyonproductivityofagricultural
fieldsandlivestock(measuredwithEUR/ha)
Input:
Landscapefieldsforagriculture
Spectrum: Positive
Developmentstage: initial
Creator: LeibnizInstituteof
AgriculturalDevelopment
Output:
Estimationofoutput(EUS/ha)foreachpolicyoninterestonborrowedcapital(IBC),intereston
equitycapital(IEC),costsavingsduetomanagerialability(MA)andfortechnologicalchange(TC)
BusinessModelforeach
commodity
6/2/2014
10
AGENTBASEDSIMULATIONFORU
BOATSSEARCHOPERATION
Description:
Thismodelistoformulateairandmaritimepatrolpatterntominimizepotencyof
infiltrationofsubmarineshipsfromenemy.
Input:
Landscapearea
Model specification of
Spectrum: Positive
Developmentstage: initial
Creator: TheAirForce
InstituteofTechnology
Output:
Estimationofsuccessofdetectinginfiltration
ofsubmarineforeachpatrolpattern.
Modelspecificationof
airplane,ship,radar
Patrolpattern
RIVERREGIONUNITin:
Prov.Banten
DKIJakarta
Prov.JawaBarat
ManagingCollaborationinCitarum River
CitarumRiverBasin:
6.080Km2
38
POLAINDUKJAWABARAT2000/ MCC2001
LectureNotesinArtificialIntelligence(LNAI),Springer
TheU.S.AirForceResearchLaboratory
POPULATION
PRESSUREcauses
deforestation
TOOMUCH:
FLOOD
ThreeClassicalProblemsofCitarumRiverbasinProblem
RUSTAM SYARIF/ 2002
UNCARECULTURE
causesrubbishalong
theriver
TOOLITTLE:
DROUGHT
RUSTAMSYARIF/ 2002
RUSTAMSYARIF/ 2002
39
Untreatment
wastedisposalfrom
industries
TOODIRTY:
DISEASE
RUSTAMSYARIF/ 2002
ActorsinvolvedinDASCitarum
Citi i
Selfish behavior (silo)create
manysocialproblems
Needcollaboration
Citiesin
downstre
ams
Textile
industries
Local
peoplein
upperstrea
Environme
Regencies
inupper
stream
40
upperstrea
ms
Peoplein
downstrea
ms
Environme
ntalists
6/2/2014
11
PartialProposalsCreateConfrontations
STOP DEFORESTATIONNOT HAPPY UPPER LOCAL GOVERNMENT
STOP ILLEGAL LODGING NOT HAPPY UPPER STREAM LOCAL PEOPLE
AUTHORITARIANGOVERNMENTNOT HAPPY COMMUNITY
SELF INTERESTS UPSTREAM SUFFER DOWNSTREAM
41
SELF INTERESTS UPSTREAM SUFFER DOWNSTREAM
DOWNSTREAMDOESNT CONTRIBUTE TO UPPER UNPROTECTEDUPSTREAM
BLAME INDUSTRIESUNEMPLOYMENT
OVERACT ENVIRONMENTALIST STAGNANT DEVELOPMENT
ATLASTTHEENVIRONMENTISMOREANDMOREDEGRADED
ANDTHEPEOPLEISTHENBEINGVICTIMS
QualitativeResearches
Citiesin
downstr
eams Textile
industrie
s
Regencie
sin
upper
Selfish behavior (silo)Createsocial
problemsNeedcollaboration
Local
peoplein
upperstre
ams
Peoplein
downstre
ams
Environm
entalists
stream
Interview
Interactionmodel
amongstakeholders:
1. Position
2. Interest
3. Dilema
4. Emotion
DramaTheory
42
Modelofstakeholderinteraction
CurrentCRF Options
Agent 1`s
position
Agent-2`s
position Options
Agent 1`s
position
Agent-2`s
position
Agent1 Agent2
NewPayoff
NewEmotion
O1 A (75) R (87)
O2 R (79) A (89)
O1 A (75) R (87)
O2 R (79) A (89)
Negotiatian
CurrentDilemma
NewEmotion
NewPerception
NextCRF
NextDilemma
Internalattributes:
Payoff/Preference
Perceptionotheragentspayoff
Bargainingstrategy
Emotion
Internalattributes:
Payoff/Preference
Perceptionotheragentspayoff
Bargainingstrategy
Emotion
43
ManagingCollaborationinCitarum
RiverBasin
Citiesin
downstr
eams Textile
industrie
s
Regencie
sin
upper
stream
Interview
Local
peoplein
upperstre
ams
Peoplein
downstre
ams
Environm
entalists
stream
Selfish behavior (silo)Createsocial
problemsNeedcollaboration
Scenariosmayhappenin
future
HappyEnding&
Tragedy
S
i
m
u
l
a
s
i
Strategytodevelop
collaboration:
Positionsnecessary
Emotion
Stakeholderwhoneed
activelytakearole
I
d
e
n
t
i
f
i
k
a
s
i
P
o
l
a
44
Interview
Interactionmodel
amongstakeholders:
1. Position
2. Interest
3. Dilema
4. Emotion
DramaTheory
6/2/2014
12
Interactionmodelamongconsumersinairlineservices
Consumers need to select airline suitable for them when
IEEETransactionsonSystems,Man,andCybernetics,30(5)
A chapterinAgentbasedapproachesinEconomicsand
SocialComplexSystemsIV,Springer
Example Example::
Browsingsosial mediavia online or
Facetofaceaskingato friendsabouttheirexperiencesinusingairline
providers.
Consumersneedtoselectairlinesuitableforthemwhen
theyneedtotravel
45
p
45 45
Facetoface SocialMedia
Commonundertanding oneachsidesexpectation
increasesservicesatisfaction
Communication
Style
CoExperience
Maybeinconsistent
Consumerperception
(Ekspectation onprovidersservice)
Providersperception
(Expectationonservice
neededbyconsumers)
Consumer
Provider
46
AdaptiveLearningModelusingGenetic
Algorithm
Interaction
among
Ratio of customerwho know provider
C t i il it among
consumers
Selection
Crossover
GA
Operators
Stopping
Criterion
Satisfied?
Present
The best
Individual
of end
population
Yes
Roulette
Wheel
Communicationstyle:
Facetofaceand
socialmedia
Customerssimilarity
47
No
Dynamically
changing
One-point
crossover
Customer
Consumer
Parameter:
Learningefficiency
Learningperformance
SimulationResults
Hypothesis:
Ifconsumersperceptionsarediversethensocialmediaisanefficientwayto
d t d id
15
20
25
30
35
40
A
v
e
ra
g
e
D
is
ta
n
c
e
D istance M atrix betw een C ustom er and Provider:Social-M edia B ased (SN S) and Face-to-face

understandprovider.
Hypothesisisaccepted
p
e
r
s
e
p
s
i
d
e
n
g
a
n
p
r
o
v
i
d
e
r
10 20 30 40 50 60 70 80 90 100
0
5
10
G eneration

Face-to-face(50 custom ers)
Social-M edia B ased (1000 custom ers)
48
Socialmedia
P
e
r
b
e
d
a
a
n
p
6/2/2014
13
ThankYou

Você também pode gostar