Analysis of a Business Environment using Burstiness Parameter: The
Case of a Grocery Shop
Andreas Ahrens
1
, Ojaras Purvinis
2
, Detlef Hartleb
1
, Jelena Zaˇsˇcerinska
3
and Diana Miceviˇcien˙e
4
1
Hochschule Wismar, University of Technology, Business and Design, Wismar, Germany
2
Kaunas University of Technology, Kaunas, Lithuania
3
Centre for Education and Innovation Research, Riga, Latvia
4
Panev˙zys University of Applied Sciences, Kaunas, Lithuania
Keywords:
Buyers’ Burstiness, Independent Event, Gap Processes, Binary Customer Behaviour, Burstiness Estimation,
Burstiness Measurement.
Abstract:
Nowadays, bursty business processes are part of our everyday life. Bursty business processes include such pro-
cesses as selling and buying, too. One of the contemporary challenges business environment has to deal with
is monitoring and controlling of burstiness in business processes. Monitoring and controlling of burstiness in
business processes often leads to the optimization of business processes. Validation of the model for analysing
buyers’ burstiness in business processes revealed the need in optimisation of the proposed model, as the elab-
orated model based on gap processes is complex for implementation, as well as for parameter estimation. For
optimization of the model for analysing buyers’ burstiness in business process, different levels of burstiness
in the process of buying are studied in this work. Different approaches to modelling buyers’ behaviours are
presented and evaluated in this work, too. The novel contribution of this work is based on the estimation of
burstiness. With the proposed solution the level of burstiness can be estimated by taking the mean value and
the standard deviation of a gap sequence into account, which always exists for a given sequence. As a practical
application, the cash register of a medium size grocery shop in Lithuania is analysed. The novelty of this paper
is given by the comparison of different approaches to measuring burstiness in real process data. Directions of
further research are proposed.
1 INTRODUCTION
Nowadays, bursty business processes or, in other
words, business environment are part of our everyday
life. Traffic flow in cities is bursty, data traffic implies
to be of a bursty nature, customers’ flow in shops are
not static, too, as shown in Fig. 1.
Bursty Business
Processes
Trac Flow Data Flow Customer Flow
Figure 1: A range of bursty business processes in everyday
life.
Bursty business processes include such processes
as selling and buying, as shown in Fig. 2. One of the
contemporary challenges business environment has
to deal with is monitoring and controlling of bursti-
ness in business processes. Monitoring and control-
ling of burstiness in business processes often leads
to the optimization of business processes. Burstiness
Bursty Business
Processes
Selling Buying
Figure 2: The inter-relationship between bursty business
processes as well as selling and buying.
has attracted a lot of attention starting with the mod-
elling of bursts or bundles of bit-errors in telecom-
munications. Such investigations have led to inten-
sive research for simulation models which are able to
take the bursty characteristic of bit-errors into account
such as (Gilbert, 1960) or (Elliott, 1963). Similar de-
pendencies can be found in data networks regarding
the characteristics of the traffic (e.g. the temporal
intervals between consecutive data packets) (Kessler
Ahrens, A., Purvinis, O., Hartleb, D., Zaš
ˇ
cerinska, J. and Micevi
ˇ
cien
˙
e, D.
Analysis of a Business Environment using Burstiness Parameter: The Case of a Grocery Shop.
DOI: 10.5220/0007977600490056
In Proceedings of the 9th International Conference on Pervasive and Embedded Computing and Communication Systems (PECCS 2019), pages 49-56
ISBN: 978-989-758-385-8
Copyright
c
2019 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
49
Table 1: Burstiness in different scientific fields.
Scientific field Phenomenon of burstiness
Telecommunications Burstiness of bit-errors in data transmission
Economics Burstiness of crises
Natural sciences Burstiness of disasters or earthquakes
Logistics Burstiness of traffic
Social media Burstiness of hot topic, keyword or event
Business Burstiness of workload
Business Burstiness of buyers
et al., 2003; Feldmann, 2000).
Furthermore, the phenomenon of burstiness was
revealed in a range of scientific fields such as eco-
nomics, natural sciences, logistics and business.
Tab. 1 demonstrates the phenomenon of burstiness in
a range of scientific fields.
In business, burstiness is based on visitor-buyer
relationship as illustrated in Fig. 3. The visitor-buyer
relationship implies binary customer behaviour such
as buying or not buying. A visitor becomes with the
probability p
e
a buyer (also referred as buyer prob-
ability) and remains with the probability (1 p
e
) a
visitor.
Visitor Buyer
p
e
(1 p
e
)
Figure 3: Visitor-Buyer Relationship.
However, these models do not take the concentra-
tion of buyers into account as highlighted Fig. 4.
x x - x x x - - x x x x x x - x x - - - - - - - - - - - -
- - - - - - - x x x x - x - - x x - - x x - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - x x x x
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - - - x x - - - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - x x x x - - - - - x - - x x -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - x x x - - - - x x x - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - - - - - x x x - - - - - - - - - - - - - - - x x - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
x x x x x - - - - - - - - - - - x x - - x x x x x x - x -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
- - - x x x x x x - - - - - - - - - - - - - - - - - - - -
Figure 4: Buyers’ burstiness (represented by ”x”) within a
sequence of shop visitor (represented by ”-”).
In (Ahrens and Zsˇcerinska, 2017) a model for
analysing buyers’ burstiness in business processes has
been presented. The model shows that the process of
buying can be described by the buyers’ probability.
However, in order to be able to describe the bursty
nature of buyers a second parameter such as the buy-
ers’ concentration is needed.
The optimization of bursty business processes re-
quires, on the one hand, appropriate simulation mod-
els and, on the other hand, algorithms for estimating
burstiness in business processes in order to be able to
optimize process systems such as queuing at the cash
register in a shop.
Optimization of business processes also depends
on a level of burstiness. Hence, an issue is the mea-
surement of burstiness. A couple of approaches to the
measurement of burstiness exist. The F (Fano) factor
as well as burstiness factor (1 α) (also referred as
parameter in the present research) are widely used to
estimate a level of burstiness (Ahrens et al., 2019b).
In this work the level of burstiness in the process
of buying is studied. As a practical application, the
cash register of a medium size grocery shop in Lithua-
nia is analysed. The proposed solution of burstiness
estimation takes the mean value and the standard de-
viation of real data into account and avoids the com-
plex estimation of distribution or density functions.
The novelty of this paper is given by the compar-
ison of different approaches to measuring burstiness
in real process data.
The remaining part of this paper is organized as
follows: Section 2 introduces the theoretical basis for
modelling buyers’ behaviour. A mathematical model
for describing buyers’ burstiness via gap processes is
presented in Section 3. The estimation of burstiness
is introduced in Section 4. The associated results of
an empirical study of a medium size grocery shop in
Lithuania are discussed in Section 5. Finally, some
concluding remarks are provided in Section 6.
PECCS 2019 - 9th International Conference on Pervasive and Embedded Computing and Communication Systems
50
2 THEORETICAL BASIS
In this section the theoretical basis for describing
business processes with independent events of buy-
ers (i.e. the buyers appear independently from each
other) is given. In general, any process including
the process of buying in which binary decisions are
made can be described by gaps as illustrated in Fig. 5
(Ahrens et al., 2019a). Once the process is modelled
- x - - x - - x - - - x x - - - - x -
2 2 3 40
Figure 5: Modelling of the buying process by gaps (a buyer
(represented by ”x”) within a sequence of non-buying visi-
tors (represented by ”-”)).
by gaps, a gap distribution function u(k) defining the
probability that a gapY between two buyers is greater
than or at least equal to a given number k, i.e.
u(k) = P(Y k) (1)
can be defined within the following boundaries
u(0) = 1 and lim
k
u(k) = 0 . (2)
Next to u(k) a gap density function v(k) defining the
probability that a gap Y between two buyers is equal
to a given number k, i. e.
v(k) = P(Y = k) (3)
can be defined. Taking v(k) into account, the function
u(k) results in
u(k) = v(k) + v(k + 1) + v(k+ 2) + · · · . (4)
For situations with independent events, i. e. buyers,
u(k) can be defined, as a function of the buyers’ prob-
ability p
e
, as follows
u(k) = (1 p
e
)
k
=
p
k
e
. (5)
Equation (5) is well-known in probability theory for
independent events and is valid for any buyer proba-
bility p
e
. Therein, the probability of non-buying visi-
tors is defined as
p
e
=
Number of Visitors - Number of Buyers
Number of Visitors
. (6)
With (5) the probability can be derived that k con-
secutive visitors are non-buying visitors.
By calculating the average gap length E(Y), the
interrelation between u(k) and p
e
becomes visible.
Here, we get
E(Y) + 1 =
1
p
e
. (7)
Calculating the sum of u(k), we receive
k=0
u(k) = u(0) +
k=1
u(k) = 1+
k=1
u(k) . (8)
With (4), the expression can be re-written as
k=0
u(k) = 1+
k=1
k· v(k) (9)
and the calculation of the average gap length E(Y)
becomes
E(Y) =
k=0
k· v(k) =
k=0
u(k) 1 . (10)
The function v(k) can be decomposed with (4) as
v(k) = u(k) u(k+ 1) . (11)
Combining (7) and (10), we get
k=0
u(k) =
1
p
e
. (12)
By taking (5) into account, (12) can be verified as fol-
lows
k=0
u(k) =
k=0
(1 p
e
)
k
=
1
1 (1 p
e
)
=
1
p
e
. (13)
Furthermore, defining a given interval n with at least
one buyer, the block buyer probability p
B
(n) is for-
mulated as
p
B
(n) = 1 (1 p
e
)
n
(14)
and can be approximated for small p
e
as follows
p
B
(n) 1 (1 n p
e
) = n p
e
(15)
with
lim
n
p
B
(n) = 1 . (16)
The block buyer probability p
B
(n) results from the
probability of having no buyers in a block of the
length n, defined as (1 p
e
)
n
.
In a double-logarithmic representation the linear
relationship between log
10
(p
B
(n)) and log
10
(n) be-
comes for p
B
(n) 1 evident. Here we get
log
10
(p
B
(n)) = log
10
(n) + log
10
(p
e
) . (17)
as it was confirmed by practical measurements (Wil-
helm, 1976).
Furthermore, the block buyer probability p
B
(n)
can be calculated by taking the distribution function
u(k) into account as shown in Fig. 6 and results in
p
B
(n) = p
e
·
n1
k=0
u(k) (18)
Together with p
B
(n) = p
e
n, derived in (15), we get
n1
k=0
u(k) = n . (19)
The searched distribution u(k) can now be obtained
iteratively
Analysis of a Business Environment using Burstiness Parameter: The Case of a Grocery Shop
51
block interval
n = 3
P(1 buyer on position 1)
=
st
p u(0)
e
P(1 buyer on position 2)
=
st
p u(1)
e
P(1 buyer on position 3)
=
st
p u(2)
e
buyer
with p
e
Figure 6: Calculating the block-buyer probability p
B
(n) using the gap distribution function u(k).
n = 1 : u(0) = 1
n = 2 : u(0) + u(1) = 2
n = 3 : u(0) + u(1) + u(2) = 3
· · · : · · · = · · ·
n : u(0) + u(1) + · · · + u(n 1) = n
and formulated as
u(k) = (k+ 1) k = 1 . (20)
In order to fulfil (2) and (12), the function u(k), de-
fined in (20), can be multiplied by
(1 p
e
)
k
e
p
e
·k
(21)
resulting in (5).
The resulting block buyers’ probability p
B
(n) is
highlighted in Fig. 7 when analysing the differences
in (15) and (18). The linear dependency between
log
10
(p
B
(n)) and log
10
(n) for p
B
(n) 1 forms a ba-
sis for buyers’ simulation model.
3 DESCRIPTION OF BURSTY
BUSINESS PROCESSES
The bursty business processes have to consider that
the block buyer probability decreases for a given n as
the buyers become more and more concentrated. By
introducing a buyers’ concentration (1 α) as shown
in (Ahrens, 2000) and (Ahrens and Zaˇsˇcerinska,
2016) the block buyer probability can be approxi-
mated as
p
B
(n) = p
e
n
α
(22)
for any interval with p
B
(n) 1. In the double-
logarithmic representation we get
log
10
(p
B
(n)) = α log
10
(n) + log
10
(p
e
) . (23)
0 1 2 3 4
-2
-1.5
-1
-0.5
0
log
10
(p
B
(n))
log
10
(n)
Theory
Approximation
Figure 7: Approximated block buyer’s probability p
B
(n) as
a function of the interval length n for (1α) = 0 at a buyer’s
probability of p
e
= 10
2
.
with the parameter α defining the gradient of the line.
Practically relevant buyers’ concentrations are in the
range of 0 < (1 α) 0.5, whereas a buyers’ concen-
tration of (1 α) = 0 describes the beforehand stud-
ied situation with independent buyers. Using (18), the
distribution function u(k) results in
n = 1 : u(0) = 1
n = 2 : u(0) + u(1) = 2
α
n = 3 : u(0) + u(1) + u(2) = 3
α
· · · : · · · = · · ·
n : u(0) + u(1) + · · · + u(n 1) = n
α
and can be calculated as
u(k) = (k + 1)
α
k
α
. (24)
In order to satisfy the condition
lim
k
k
κ=0
u(κ) =
1
p
e
(25)
PECCS 2019 - 9th International Conference on Pervasive and Embedded Computing and Communication Systems
52
(24) can be multiplied with the asymptote e
β·k
with
lim
k
e
β·k
= 0 . (26)
By multiplying u(k) with the factor e
β·k
, the param-
eter β has to be calculated in order to fulfil the condi-
tion
k=0
[(k+ 1)
α
k
α
] · e
β·k
=
1
p
e
. (27)
Taking the series expansion of the expression
(k+ k)
α
= k
α
(1+
α
k
k+ ...) (28)
into account, the expression (24) simplifies with k =
1 to
(k+ 1)
α
k
α
α · k
α1
. (29)
Using the integral instead of the sum, the following
equation has to be solved in order to determine the
parameter β. Here we get
U = α
Z
0
k
α1
e
β·k
dk =
α Γ(α)
β
α
, (30)
with the parameter Γ(·) describing the Gamma func-
tion. With the approximation
α Γ(α) 1 (31)
we get
k=0
[(k+ 1)
α
k
α
] · e
β·k
=
1
β
α
. (32)
Together with (27) the following approximation for
parameter β has been found
p
e
β
α
. (33)
For bursty buying processes, the buyers’ gap distribu-
tion function results in
u(k) =
k=0
[(k+ 1)
α
k
α
] · e
β·k
. (34)
For independent buyers, i. e. α = 1, the parameter β
equals the buyer probability p
e
as derivedin section 2.
Fig. 8 demonstrates the buyers’ gap distribution
functions. With increasing buyers’ concentration
(1 α), the appearance of gaps of shorter lengths in-
creases whereas at the same time the probability of
longer gaps decreases.
Fig. 9 shows the calculated block buyers’ proba-
bilities p
B
(n) as a function of the interval length n for
different parameters of the (1 α) at buyers’ prob-
ability of p
e
= 10
2
. With increasing buyers’ con-
centration, the buyers appear more and more concen-
trated and block buyer’s probability p
B
(n) decreases
for a given n as stated before. Furthermore, the lin-
ear dependence between log
10
(n) and log
10
(p
B
(n))
becomes obvious for small n.
0
0.2
0.4
0.6
Probability
gap length
(1 α) = 0.2
(1 α) = 0.1
(1 α) = 0.0
0 ... 10
11 ... 20
21 ... 30
Figure 8: Probability of different gap length for different
parameters of the (1 α) at a buyer’s probability of p
e
=
10
2
.
0 1 2 3 4
-2
-1.5
-1
-0.5
0
log
10
(p
B
(n))
log
10
(n)
(1 α) = 0.0
(1 α) = 0.2
(1 α) = 0.5
Figure 9: Approximated block buyer’s probability p
B
(n) as
a function of the interval length n for different parameters
of the (1 α) at a buyer’s probability of p
e
= 10
2
.
4 ESTIMATION OF BURSTINESS
The practical evaluation of bursty buying processes
requires the calculation of the level of burstiness. A
possible approach can be defined by the estimation of
the buyers’ concentration (1 α) as shown in (Ahrens
and Zaˇsˇcerinska, 2017). For this, the gap density
function v(k) has to be analysed. Analysing the prob-
ability that after a buyer, in the distance of zero an-
other buyer appears, i.e. v(0) = u(0) u(1), the buy-
ers’ concentration (1 α) can be analysed by taking
(34) into account. Here we ge with u(0) = 1
v(0) = 1 u(1) = 1
h
(2
α
1) e
β
i
. (35)
With the assumption of small beta the expression can
be simplified as follows
e
β
1 for β 1 , (36)
and the parameter v(0) simplifies to
v(0) 2 2
α
. (37)
From this equation, the buyers’ concentration (1 α)
is estimated as
(1 α) 1 log
2
[2 v(0)] . (38)
Analysis of a Business Environment using Burstiness Parameter: The Case of a Grocery Shop
53
However, it requires the calculation of the gap density
function v(k). In comparison to (38), Goh & Barabasi
(Goh and Barab´asi, 2008) provided an alternative so-
lution for estimating burstiness in business processes.
In (Goh and Barab´asi, 2008) burstiness is defined by
taking the mean value m (average gap length or aver-
age length of a time interval between two buyers) as
well as the standard deviation σ of the length of time
intervals or gaps into account. The definition by Goh
& Barabasi (Goh and Barab´asi, 2008) results in
B =
σ m
σ+ m
. (39)
with 1 B 1.
Goh & Barabasi pointed out that B = 1 corre-
sponds to a bursty environment whereas B = 0 is re-
ferred to a neutral environment. Regular (periodic)
signals are described by negative parameters of B.
Analysing a buying process with independent buyers,
i. e. (1 α) = 0, the gap distribution function u(k)
results in
u(k) = e
p
e
·k
. (40)
Taking the gap density function v(k) = u(k) u(k+1)
into account, the mean value m = E(Y) can be calcu-
lated as follows
m =
k=0
kv(k) =
k=0
k(e
p
e
·k
e
p
e
·(k+1)
) (41)
and results in
m =
e
p
e
1 e
p
e
. (42)
Together with the standard deviation
σ =
q
E(Y
2
) m
2
=
e
p
e
/2
1 e
p
e
(43)
the parameter B results in
B =
σ m
σ+ m
=
e
p
e
/2
1
e
p
e
/2
+ 1
. (44)
Fig. 10 shows the dependence of the parameter B
on the buyers’ probability p
e
. As shown by Goh &
Barabasi the parameter B is close to zero indicating
the independence of the buyers. However, the depen-
dence of the parameter B on the buyers’ probability
shows the weakness of the burstines definition as in-
dependent buyers’ scenarios are solely described by
the buyer probability. On the other hand the param-
eter B can be easily calculated for an empirically ob-
tained gap sequence.
Taking different parameters of the buyers’concen-
tration (1 α) into account, Fig. 11 shows the ob-
tained values for the parameter B. As obtained by
computer simulations, the parameter B can be used
0.02 0.04 0.06 0.08 0.1
10
-4
10
-3
10
-2
p
e
B
Figure 10: Dependence of the parameter B on the buyer’s
probability p
e
for independent buyers.
as an indicator regarding the level of burstiness. Ac-
cording to Fig. 11 a rough estimation leads to the fol-
lowing condition
B (1 α) . (45)
Unfortunately, the plot depicted in Fig. 11 shows that
the parameter B depends on the buyers’ probability p
e
and buyers’ concentration (1 α), i.e.
B = B(p
e
,(1 α))
is used as an indicator for the expected buyers’ con-
centration.
0.02 0.04 0.06 0.08 0.1
10
-4
10
-3
10
-2
10
-1
10
0
p
e
B
(1 α) = 0.0
(1 α) = 0.2
(1 α) = 0.5
Figure 11: Dependence of the parameter B on the buyer’s
probability p
e
for different parameters of the buyers’ con-
centration (1 α).
5 PRACTICAL APPLICATION
In real world business processes, the probability p
e
of visitor to buy a good as well as the correlation be-
tween buyers, described by the buyers’ concentration
(1 α), may be not available. The solution is to use
PECCS 2019 - 9th International Conference on Pervasive and Embedded Computing and Communication Systems
54
a statistical approach to estimate the buyers’ concen-
tration (1 α).
In this section the service of the buyers at the cash
register as a practical example of bursty processes is
analysed. In this example, the duration of the service
of the buyers at the cash register of a medium size
grocery shop in Lithuania is studied. The cash regis-
ter data collected contain the data about the operation
time, the amount of goods purchased, their codes, and
the prices paid by each buyer. The data collection was
carried out in June 2018. At that time 2575 buyers
were served.
Unfortunately, the cash registers do not record the
start time of the operation. Therefore, the service
duration time was not available from the database.
To cope with this problem we observed buyers’ ser-
vice durations with different quantities of goods (see
Tab. 2). It appeared, that the service duration t
s
de-
pends not only on the quantity of the goods, but also
on the type of goods, individual characteristics of the
buyer and other random factors, i. e. the dependence
is statistical.
Table 2: Duration of the service at the cash register.
Amount of Goods Service Time
g t
s
(ins)
3 44
1 18
10 30
1 11
18 61
1 37
The correlation coefficient between g and t
s
equals
0,72 and the regression equation is given by
t
s
= 1,9g+ 22,8 . (46)
The equation yields that for one good about 1,9 sec-
onds and additionally about 22,8 seconds for each
buyer are required.
Knowing the quantity of goods and (46), the start
and end times of each buyer can be calculated. This
allows us to analyse the free time between two buyers’
service, if any. When it appeared that there was no
free time interval between two or several buyers’ ser-
vice, then the sequential service times were merged
into one continuous service time interval.
Therefore, it was possible to investigate service
duration times and time gaps between successive buy-
ers. It appeared that the average service duration is
37 seconds and the most frequent duration takes 32
seconds. The distribution of service duration time is
given in Fig. 12.
Similarly, the durationof free time intervals (gaps)
can be processed. The average free time interval (a
0 20 40 60 80
0
0.2
0.4
Probability
t(ins)
Figure 12: Distribution of service duration times.
gap) equals m = 234 s, i.e. 4 minutes and the stan-
dard deviation σ = 620 s. The histogram is given in
Fig. 13.
0 100 200 300 400 500 600
0
0.1
0.2
Probability
t(ins)
Figure 13: Distribution of free times of cash register
(grouped).
This histogram is different from the histogram of
service times. The frequencies of free times (gaps
of various lengths) are constantly decreasing while
it is not true for histogram of service times given in
Fig. 12.
The small mode of free time (gap lengths) and the
histogram of free times testifies that the time gaps
between services usually are short. The histogram
of free times’ durations up to 30 seconds, i. e. half
minute, reveals that these durations are distributed
quite similarly (see Fig. 14.). One of the measures
0 10 20 30
0
0.05
0.1
Probability
t(ins)
Figure 14: Distribution of free time duration up to 30 sec-
onds.
of burstiness is the parameter B defined in (39). Ap-
plying this formula to free times (gaps), it yields that
B = 0,45. Therefore, the free times between services
are quite bursty. On the other hand, the burstiness
level of free times up to 30 seconds states that this
process is close to neutral. The high burstiness of free
times during the whole month was determined by the
longer free time at the beginning and end of the shop
open time.
Analysis of a Business Environment using Burstiness Parameter: The Case of a Grocery Shop
55
6 CONCLUSIONS
In this work, on the example of the cash register of a
medium size grocery shop in Lithuania, different ap-
proaches to estimation burstiness are presented and
analysed. The proposed solution of burstiness esti-
mation takes the mean value and the standard devia-
tion into account and avoids the complex estimation
of distribution or density functions.
The discussed probabilistic models and their ap-
proximations of business processes can be evaluated
by the burstiness parameter B. It revealed, the bursti-
ness is positive, i. e. between neutral and bursty pro-
cess in the investigated case of a grocery shop in
Lithuania.
In real world business processes, the probability
p
e
of visitor to buy a good as well as the buyers’ con-
centration (1 α) may be not available. Nevertheless,
it is possible to process statistically the cash register
data. Usually the cash equipment just registers one
time moment of the service of the buyer and number
of goods and their codes in the basket, but not the
service duration. Therefore, the shop’s database does
not contain lengths of busy intervals and lengths of
free time intervals. The solution of the problem is an
additional observation of the cashier’s work, the reg-
istration of the number of goods in the buyer’s basket
and the service time of the basket. Then the regres-
sion equation between the number of goods in the
basket and service duration was derived. Using this
equation, it becomes possible to estimate the service
time lengths, to compute the free times and to apply
statistical analysis including calculation of burstiness
parameter.
Our plan on the future research is to investigate the
interrelationship between business process and visitor
decisions influenced by the behaviour of other visitors
and buyers.
ACKNOWLEDGEMENT
The authors of the present paper would like to thank
the grocery shop in Lithuania for supporting the mea-
surement campaign and providing the cash register
data.
REFERENCES
Ahrens, A. (2000). A new digital channel model suitable
for the simulation and evaluation of channel error ef-
fects. In Colloquium on Speech Coding Algorithms
for Radio Channels, London (UK).
Ahrens, A., Purvinis, O., and Zaˇsˇcerinska, J. (2019a).
Gap Distributions for Analysing Buyer Behaviour in
Agent-Based Simulation. In International Conference
on Sensor Networks (Sensornets), Prague (Czech Re-
public).
Ahrens, A., Purvinis, O., Zaˇsˇcerinska, J., Miceviˇciene, D.,
and Tautkus, A. (2019b). Burstiness Management for
Smart, Sustainable and Inclusive Growth: Emerging
Research and Opportunities. IGI Global.
Ahrens, A. and Zaˇsˇcerinska, J. (2016). Gap Processes for
Analysing Buyers’ Burstiness in E-Business Process.
In International Conference on e-Business (ICE-B),
pages 8–13, Lisbon (Portugal).
Ahrens, A. and Zaˇsˇcerinska, J. (2017). Analysing Buyers
Burstiness in E-Business: Parameter Estimation and
Practical Applications. In International Conference
on e-Business (ICE-B), Madrid (Spain).
Elliott, E. O. (1963). Estimates of Error Rates for Codes on
Burst-Noise Channels. Bell System Technical Journal,
42(5):1977–1997.
Feldmann, A. (2000). Characteristics of TCP Connection
Arrivals. In Park, K. and Willinger, W., editors, Self-
similar Network Traffic and Performance Evaluation,
chapter 15, pages 367–399. Wiley.
Gilbert, E. N. (1960). Capacity of a Burst-Noise Channel.
Bell System Technical Journal, 39:1253–1265.
Goh, K.-I. and Barab´asi, A.-L. (2008). Burstiness and
Memory in Complex Systems. Exploring the Fron-
tiers of Physics (EPL), 81(4):48002.
Kessler, T., Ahrens, A., C., L., and Melzer, H.-D. (2003).
Modelling of connection arrivals in Ethernet-based
data networks. In 4rd International Conference on
Information, Communications and Signal Processing
and 4th IEEE Pacific-Rim Conference on Multimedia
(ICICS-PCM), page 3B6.6, Singapore (Republic of
Singapore).
Wilhelm, H. (1976). Daten¨ubertragung (in German).
Milit¨arverlag, Berlin.
PECCS 2019 - 9th International Conference on Pervasive and Embedded Computing and Communication Systems
56