SSCM 4833

Di screte Event Si mulati on

 INPUT
MODELLING
1 INTRODUCTION
 CATEGORY OF
DATA

STRUCTURE OPERATION NUMERICAL

DATA DATA DATA

• Components in the • Data that represents • Gives quantitative

systems the operation of a information about a
• Interaction between system system.
the components • e.g. the entity’s path • E.g. customers’ time
• The components – data, queuing arrival, service time,
entity, source, discipline, procedure and breakdown time.
location in the for an entity to go • Can be collected from
system. through, and the observation to the
• Can be collected worker schedule. system.
through observation, • Can be collected from • Do statistical analysis
interview, the observation of the to determine the
questionnaire, and system. distribution and
system research. suitability of the data.
DATA TYPE

Qualitative Quantitative

• Data in discreet or
• Numerical data or score continuous form.
without quantitative • Discreet data – obtained
value. through counting, must
• E.g. number to represent have exact value e.g.
the choice for gender number of processed
either 1 (male) or 2 item, number of
(female). 1 and 2 here customers arrived.
does not have • Continuous data –
quantitative value. observed data in a range
and not taking and exact
value e.g. weight, height,
service time, velocity.
DEVELOPING A
USEFUL M ODEL
FOR INPUT DATA

Identify a
Collect Choose Goodness-
probability
data parameters of-fit test
distribution
2 DATA
COLLECTION
 DATA COLLECTION

 The m ost im por t an t an d dif f icult pr ocess in sim ulat ion .

 Bef or e ’f it t in g’ t he dat a occur s, t her e n eeds t o be accur at e
an d r elev an t in put dat a av ailable.
 Com m on pit f alls in dat a collect ion :
 St ale dat a
 Un expect ed dat a
 Tim e- v ar y in g dat a
 EXAM PLE:
DENTAL CLINIC

Pat ien t s com in g t o a den t al clin ic n eed t o r egist er at t he

coun t er . Next , t hey hav e t o w ait t heir t ur n s in t he w ait in g
r oom . They ar e t r eat ed based on t he queuin g sy st em f ir st - in -
f ir st - out (FIFO). Af t er t r eat m en t , pat ien t s go t o t he pay m en t
coun t er t o get t he pr escr ipt ion , m ade pay m en t an d plan f or
n ext appoin t if n ecessar y . Figur e below descr ibes t he f low of
t he sy st em .

Registration Waiting Treatment Payment

ARRIVAL LEAVING
counter Room Room counter
3 DATA
ANALYSIS
 DATA ANALYSIS

 Collect ed dat a m ust be an aly sed in or der t o m odel t he in put

dat a.
 En sur e dat a can r epr esen t t he m odel t o be con st r uct ed.
 Can use sof t w ar e such as
 SPSS
 St at :Fit
 Usin g sof t w ar e helps t o f in d suit able st at ist ical dist r ibut ion
f or t he dat a.

Choosing
Determine
the goodness-
the
probability of-fit tests,
parameters
distribution
 THE HISTOGRAM

 Div ide t he r an ge of t he dat a in t o in t er v als.

 Fin d t he f r equen cy of occur r en ces w it hin each in t er v al
 Plot t he f r equen cies on t he v er t ical axis.
 Exam ple:
The dat a in t he f ollow in g t able show t he ser v ice t im e f or 100
cust om er s w ho ar e pay in g at t he coun t er of a phar m acy .
Service Time (min) Frequency Service Time Frequency
(min)
[0,2) 15 [12,14) 7
[2,4) 11 [14,16) 4
[4,6) 19 [16,18) 5
[6,8) 16 [18,20) 3
[8,10) 8 [20,22) 4
[10,12) 8 TOTAL 100
 Can you guess the
probability
distribution that
THE HISTOGRAM match with the
histogram?

Frequency of Service Time

20
18
16
14
12
10
8
6
4
2
0
2 4 6 8 10 12 14 16 18 20 22
 THE PROBABILITY
DISTRIBUTION

Discrete Distribution Continuous Distribution

Describe random Describe random phenomena in
phenomena in which only which the variable of interest can
integer values can occur. take on any value in some interval

For example: the time to failure or

For example: demands for
the length of a rod
inventory items.

Uniform distribution
Bernoulli trials & Bernoulli
distribution
Exponential Distribution

Binomial Distribution
Erlang Distribution

Geometric & Negative

Binomial Distribution Gamma, Normal, Weibull,
Triangular, Lognormal, Beta
Distribution
Poisson Distribution
YOUR TASK!

Fin d t h e in f or m at ion on t h e f ollow in g

pr obabilit y dist r ibut ion such as t h e
par am et er s in v olv ed an d t h e gr aph of
t h e dist r ibut ion . Giv e t h e exam ples
r elat ed t o t h e sim ulat ion m odelin gs:

Bin om ial, Un if or m , Poisson

Nor m al, Expon en t ial
 SIM ULATION &
STATISTICAL
M ODEL

Inter-arrival, service time

Arrivals:- machine breakdowns, jobs coming

into job shop, units being assembled

Queuing
systems

Service time completely random – use

exponential distribution

Service time constant with random variability –

use normal distribution
 SIM ULATION &
STATISTICAL
M ODEL

Random variables:- demand per order per time period,

time between demand, lead time (between placing &
receiving orders)

Inventory & Lead time distribution can often be fitted fairly well
Supply Chain by a gamma distribution.
Model

Demand patterns can be satisfied by geometric,

Poisson and negative binomial distributions.
4 PARAM ETER
ESTIM ATION
 PARAM ETER
ESTIM ATION

DISTRIBUTION PARAMETER SUGGESTED ESTIMATOR

Poisson α 𝛼𝛼� = 𝑥𝑥̅

Exponential λ 1
λ� =
𝑥𝑥̅
Uniform (0,b) b 𝑛𝑛 + 1
b� = [maksimum (x)]
𝑛𝑛
Normal μ,σ2 μ = 𝑥𝑥̅
�
�2 = 𝑆𝑆 2
σ
 EXAM PLE

 The dat a in t he f ollow in g t able show s t he t im e t aken by

st uden t s t o com plet e a r egist r at ion pr ocess on a com put er .
Registration time (min) Frequency  As s u m in g t h e d a t a follow s
[0,2) 11
e xp on e n t ia l d is t r ibu t ion w it h
p a r a m e t e r λ.
[2,4) 10 1
 Th e e s t im a t or for λ is λ� = .
[4,6) 8 𝑥𝑥̅
[6,8) 6
[8,10) 5 ∑𝑘𝑘𝑗𝑗=1 𝑓𝑓𝑗𝑗 𝑚𝑚𝑗𝑗 298
𝑥𝑥̅ = = = 5.96
[10,12) 4 𝑛𝑛 50
[12,14) 3
1
[14,16) 2 λ� = = 0.168
5.96
[16,18) 1
5 GOODNESS-
OF-FIT TEST
 THE GOODNESS-
OF-FIT TEST

 Pr ov ide helpf ul guidan ce f or ev aluat in g t he suit abilit y of a

pot en t ial in put m odel.
 HOWEVER, in a r eal applicat ion t her e is n o sin gle cor r ect
dist r ibut ion so y ou should n ot be a slav e t o t he v er dict of
such a t est .
 The ef f ect of sam ple size in goodn ess- of - f it t est s:
 If v er y lit t le dat a ar e av ailable, t hen a goodn ess- of - f it
t est is un likely t o r eject an y can didat e dist r ibut ion s.
 If a lot of dat a ar e av ailable, a goodn ess- of - f it t est w ill
likely r eject all can didat e dist r ibut ion s.
 Thr ee m et hod f or goodn ess- of - f it t est
 Chi- squar e t est
 Kolm ogor ov - Sm ir n ov t est
 An der son - Dar lin g t est
THE CHI-SQUARE
TEST

HYPOTHESIS
01 STATEM ENT

02 CALCULATE THE
χ2 VALUE

TEST RESULT AND

03 CONCLUSION

DATA INPUT
04
 HYPOTHESIS
STATEM ENT

 The hy pot hesis is based on t he chose pr obabilit y

dist r ibut ion w it h t he est im at ed par am et er v alue.

H 0 = The r an dom v ar iable, X con f or m s t o t he dist r ibut ion al

assum pt ion w it h t he par am et er (s) giv en by t he par am et er
est im at e(s).

H 1 = The r an dom v ar iable X does n ot con f or m .

 CALCULATE THE χ2
VALUE

 The t est st at ist ics is giv en by :

𝑘𝑘
(𝑂𝑂𝑗𝑗 − 𝐸𝐸𝑗𝑗 )2
𝜒𝜒02 =�
𝑗𝑗=1
𝐸𝐸𝑗𝑗

Wher e
𝑂𝑂𝑗𝑗 = the observed frequency in the 𝑖𝑖th class intervals
𝐸𝐸𝑗𝑗 = the expected frequency in that class interval
 TEST RESULTS
AND CONCLUSION

 Based on t he f ollow in g con dit ion s

 Accept H 0 if t he t est v alue is sm aller or equal t o t he cr it ical
v alue. 2 2
χ0 ≤ χ𝛼𝛼 ,𝑣𝑣

 Reject H 0 if t est v alue is lar ger t han t he cr it ical v alue.