Course utline

eiaon Thecy defintiond, decisicn strede

minimx cetenons, Ukelihood
prindple,
Prier. f Postencr dishnbubien
Orimunn shrategies
Estiction in decisicn framewck.
Came Thecns , dehinition , foo -person
games and ito expensiens
Decisicn of risk fun ctions, optimiaaton
technique:
Rejerence
>Benardo 3-M e smta AFM (ig9+)
Bayesian Thecny, wilen , Neyork
>Carin, BP and Lavs iA Ci446) Bauks and
mpincad Biyes melcds

Jntnducticn
befnition
becision Theony
t is the shudy
Crguni2ation qents make clicices clen
 Deusíen
hraty 6 <
is Skicntaitern
aties
tTCng A
Same Citeric becsSomeanorng
be becision
about
are
UnertaintyAubablihycndtheoy
tesion maximiaing
esçectedis ciion
docisicn
htenaas COmmon ien [Link]
deusicn atemative t cnteria. themakers
quantty
ntd shatagy fheccy
a
gui te potentd anises
onterts
ncus decision expeted deisicn
tbeie
shdegies oriodude:
unteria decison prui may
ohere
saisying censequenEs Kus
des infomation
complete
tack
cat lenten maker pan ublity
makersYaríous wncertobty
nbabilihes
to minimiing
certain
in índeciicsn
may choase t to
sion be chocsirg eir of
oshats doices atomes model

Aternatves a.
notchere 4Uncer
tanty
maximixs
act
heir he Deision
ansiders numeícalpr«{crenes
theoy based
Or hese have Aeng [Link]
mes becision
b CUtcmakers juncertainty
cuteemg
Many akisiong m chen
degsicoacions ublihy providesa fameucrk
naun
docipoblisnievcves
cmsn toced
cith makEr Hne valtuncbon
ue
availcublg decisions
kueioverat
r ecable
pieerenes
decisicn whwkan
tiny t GS
mut cach wlich SumEs
epede tiy dchcea o makers
mibple
dfferent
actcnsane e possibleass'repeanted
usuaiy
gns Kaat
kideasi
e cn peAe
decisien itenctves
unertainihy. ateratv maker [Link]

and
 mmigi2es JteNínentsdeisico-
shrates
Moxinmox
oydes assecited Expeced 2ticn Ciganí maiinogent after Decisicn
beision detns
Cc
fronteLocrkfor megene
businas,
tincduding
fields
is
daisicn.
nrking censidenn a
peuides
risk- sJetinq
secking a mcknq
ihe
uslty
anti enhy
sudn
endussion
citn an LA)
maxlmum atcene, cip
ditferent impeuoq
prcELS and a
syslemabi
understand1g
Hhe ted Pgrmme,
cemputei
rarpenibe dtterent and
or
aiernatives coniderinqaverage i0 FsseuoiggS
o
solution
apprachecible pesrible alternaties (cmplex and ecnemie
rihud s,
otennes. er vigoreus
tat the reached and
payof
probabities cr
ophans unerain

multiple
decisicn
makers
cutcmes Jame ditferent RiUhostenity
invclves
outrcnnes cRiskand
baision n sk outomes
Deison happiness ublity1heis |Aecrding tecry
altern
ctive
rabond
hecny, ome invces
thay on ihiaticns
Theory
Cire the where cases,
is itubions qssociied eciion to
intuenceddeasion decision distingwstes Hne
widely deisíen ehere are
maker
mpeced
ct whi'e uhere
uncertanty
knceen
applied by te cf
Theony probabi ith
ldd
interacticns cthersand probabilihes nisk each ulity
in extend
lites dncese
3. oher
unertty reeg,a
Mne
bh è to cf paside
 best cutcame. Expecded vae stateg
Problcm fonnton. lnvoles sdechinq the aternatie with tn
t is cnuaal step in decsion analyi highet apeácd valie
and invchves stuconng he decisicn prblem, b Considers both he probabilihes ond paues
payctts
deñninq ckjectives, identiging alkerncthives? aseiatedith duferent tudcemer
understandgg he dsip bin diterent clemerts Maxí mum sbatgy
Jntuence diagam, payctf iables aad the Jnyoives seecing tne alternatve thdt merimí2es
deisicn tees are tocls used in probiem Ahe minímym pasible payott
fomdtien. te represent Cn anyaed deusico |k fs a nisk -Averse apooach
Piublem Minimax Appyect Reqret sheteay
diagiam volves sucing tne aiteYncive tiad mioimies
Intuene diagram sually representa fhe he maximum regret chere regret is Hhe
difference btn the best pasribe outome and
Hetuip and dependenies amang van'eus the actuad outtom e
lemcnts n deis ion Prblem.
cropcrcnta intuence dicgrem Heuristics
DDeaision rode: aresimple practica nies er strategies
Kepresent deision pcints ohere decsion madeers Hhad inividuads use to make decisiens quidely
cCese aneng different cterntive s and wih less cagnitve efort
) Chance rode. Sabsfiung
Fopiesent uncertin foctos thet t s a decision strateqy chee the decwon
ioiuencE PUtome. maker seek a scluton Hhat mecb a satsactry
Ct

tbesure Yatner Kan cptinmizlng fa ë ahslte

 temtives
stateSbecision a
Ndnnçie cumn ditteentcraníreandprsent
Rcoiepresenting
values tprrsenting payetts Ihes
assoiatect
decis cncng
Faclitatetanty
diaginns
intucoc
Benets
atcnes
Componnts ío
nature Cf decisicn io
asiicted
vidues. citernativesand
representing
cmmunitation
thein
makers
diferent for_
cl [Link] ft each amplex
and
representinghe state
ombinotícn
pasible cphons
decisic n the
undestand1ogdeision
ofcutcome
ntirg.
stuctures
ef
f

NroHt (Fvourable diInguene

agam Proftsmattket
nodeforIntuene launch Cons
ider 0
nodes Repaent
decis
icns Utility
nodeud
for
diagrams EXamplae
business
fevourable NOt launch,
becision.
produc:
ccnditon de
not might decísicn
(Gavourable ,
launch
Pretit aindudei
launchlDnot utilihy
, on
chance
nde decsion whether
oirablefervNot
or
to
 "Pitceed" a
Eochofon
tomplehon"tm t e Uinthge éndpoints cutome4Aruus Represent
Chance
event Node Kepesent
unertin made Coperer cutcemes
potetialdecisico
Termina
Nodes Branche s Decision
Node. D
asicaated miglut
Preious enne decision
reremngrep aiternative
illastt
'Delay cting
wth
ncdes intspo s,
cost af ho2) unertainites
the
heexampe, fnal shocing here
'blaued deqsicn
euible outcome
cuoice qi
to
a poibe and
o oub
outome degsien
plehionm
"

ly Tre decisicn
beisier
SCCnant D
clearty Beneh Decision osts
Htemohies
The
reptesentingrepresenting Concider
naie Of
alterntives
belay Prceed payc#
altctnatve in Suows ot
companng a
ent Dayalt preiet deble "Predprjecd
he
Table cemplehicn miqut mangement
consequences
icn and
otPoyf cr
Ccast Ccost) oftPay indude time"Ddlay"
On
slecbing Nceture
tates q time
em1
and vadue or
and dewtcih
decisicn
the ditterent icn
last) Ccast) belqyc asaiated "belagt state
dtng bet
 Chance
Dode teimigl
briraomen On tbime Catl paic)
fnhdec r *Proend
implics its con cer ned belagnd Ccetpyet)
1 Cn tine
ith peeas maing decisien [Link] extensio Lbeay Termind node Ccei onyi )
to stahahca theony índudes ceisien thecy degsion ->
Delayed Ciatl pagert)
node Ghane
|mahips in he presence oF stotihal knaoledae hode
which purides into uhere there

he dassld approa ch to bí faclitate the ue

makiog dtcrence am ang Dhreides visua reprrsentation f the deisicn
unknoen quantities making praes
Oher elevGnt into índudes that ef the pasible YHalps in calalating epecked value s nd
(AsEqUEnces ewch is quanbhed by tar and assessinq ie best decis icn chategy
he prcr iroton. a)In corpcrates prebilihes o acount Aor
1he use Ct Bqye ian anaysis ancerkuiohes
cunijrcebon froudes
fer bilding and o ing
Ihe tods cllectively and decisicn makars
in undertanciag deisicn preblems uibinately
a
methods ked ithenselves to
leading t moreinermec and chen
applicctcns ond ccmputahiond and an dyte deis ien- making
atvaas
Th pnn pie þ- Maing
 he difeent crops ae derent attomativR dedding the bet
íven altem atives
Gct the
acts. Hp has te decide any one f in doion mak ng
Ahe crp (-e TCe, cahot, coto n etc. proces 8oneraily he qolacing steps
Does has to deado hom Hhis invclved;
diferent upe and has to select de bart biterent ataltem atves
act wtuch coill provde makimum qun. available for the probkn
a. Proper undernstapeing to he proble
Naure or Event abott chich deciscns to be takeo
the seleon of any act depend upen otenal a. The ehimate qain I las drererent
cich in contol any deasion aternativeL
maker Tuesc stats at nature depend Hpon E. 1he choice Ct preper ater ative om
many actots availabte atern atives
In a ccunhy Hare may be heay run, moderte Jmplemeraton ef he selected act'
rlin cY nc [Link] are difrerent states of
nture
s'milaty, in deci ding beut the prduchoa decascr moki ng Urder dgerent ent
3 ite ms, he monNfa cturer has Ac- y a manufa dure has to make
to estimate the demard cendrton g the daision abet the prnductfon c
of the tems XY2: hen these item ane
1he ditfeani conditcns dencnd dteent diferent ads or Aategies
stats ef nature.
to kide abt crep to be plnted.
eo be ryCI he Aterncctves
bice condieo
dffcrent
tolciuif undey
he
Cctter ederate
diverent
trbie
bove cttes

table.i9 ccton ,

300. Ioco Aiternatuea

(Acts) C
Rice
and
Nature
staes
bajai
(200cotton

Medscte
|000

nontdepe loo0 booBayai

Krleosn

Its monitig
gain he sttes Heche
he Hoceerinfomctíc
by
fom nature
AatsPayofs
PaycA cnd eshitertSUrey n
ich stimate he
Puy
ë or
(an
foleoring matn'x rom f. by doas
nature.
and reason
be kncunCs
diferent Iiber guess
Probab not
ned heanangemEnu.
Can be
usttion. ccoffk
andact islass Or hese have
stetes he
acreper mtnxpayoft from
perfee
fom understeoe or
ats
diteren
e
market
ncture bition
con
rops
 before monston
s not Passible
teaveu oill rain hecvilu or nt
om pat
Example
Kuppore manafucturer of buclkets "t Rs-go per can ct mou make
nchction, wat
bucket and eanc Rs.a per buck et He con. yell looc run aod probabihes for dirent hakes
buet dqy he dei ion m makes eshimate a
On he cther hand, s he pruducer a supenor
Gucality packet sel at tORs
and
fom his pat exerene
eadh ha Qam R4 per btket hd he ohers oy by saetific metiods . hus
Can sel ony 6o0 pkb a suerior qualihy iho probcbilifies cf ditkrent atde
We have doade knoo te ypo a t
ratyre qre cnly ashimater.
manufrcturer ldd o due

for
becistor
of

supior quatity Pkt, he Can to al dcisicn making [Link] duien

for
Boox4
maker has to
out af a no
qvailable atematies
As the P t on superior quduty
heroare a sihticn in hich a tajon
mere, he sldd prrtuce phts ek uporior taken;
uavty maker is requred to be
serteinty
@þecsiorn
csion meing
9. Deisicn making undcr uncttan ity maet
om he result Can
ersily kneen, é deasion
Haal
kking deeisions eede he best aien
qdion qre unartin, he process has to
 t
sclected.v minimum
cbHained. Pay é are |Eventuttes
Ncho a
chs A
cccrdinq Moki-
Ofprreíple
Mes
min PayBs2è
{t
act
diteren to principe
ternatve
tus
t
Ata paycffs
nCimumvalue it principle ad
pyetsaungam AclAlkernahyeA
S
terncdves Aa Aa
a
cmeupends
ced betnd
olich ditfeênts l4F
he
spebing Higl1ast
enevue -8 l6
qrairug
e to
Aa
Ad -it. lo -4

qctnaxium matrixand Jeadg

according )Maxi-min
HoriMoxi-m
cehu) an
) urertanty
piocple
laplace Mehods knowarneotchen O. he
olution Pnipiple
Maxi- m becision
deaded poblen
to
vadue pinciple efprobabilihe
Ku'thse, CAn

to act
friruple
pesimiahu
Pnnipledeien maling. af
decision
be
act n
ax-mn lo
Rgarded mininam pbm mak'ng undestudiodmaling
tecrvds
pprnci iste diterentstat ew
when uncitainity
hnciple under under
pauct pr
art Q
nct heod
breup.
deided. The
value s Hne best Aot according to
Maxi-max principle