Mobile Broadband Traffic Forecasts in Korea
Chanwoo Cho and Sungjoo Lee
Department of Industrial and Information Systems Engineering, Ajou University,
San5, Woncheon-dong, Yeongtong-gu, Suwon, Korea
Keywords: Mobile Traffic, Forecast, Smart-Phone, Smart-TV, PC, Patterns of Use.
Abstract: During many years, the dominant traffic in mobile broadband networks was voice. However, with the
introduction of diverse mobile broadband equipment, the situation has changed. Since mobile broadband
devices can allow users to access information instant and connect to web quickly, the mobile world has been
revolutionized, where global mobile data traffic has been increasing dramatically. And the changes in the
patterns of usage for mobile devices have started to cause traffic jams on the mobile broadband networks.
As a result, forecasting the future traffic needs is in urgent need to provide high-quality mobile broadband
services. To meet this need, this research aims to suggest a new forecasting method for future mobile
broadband traffic. For the purpose, three-round Delphi survey was conducted to identify devices and
applications that would affect in the future mobile broadband traffic, and their expected growth rates of
users and changes in the patterns of use for each device. Then the total amount of mobile broadband traffic
was forecasted based on survey results. The research results are expected to provide the basic research data
for a further study.
1 INTRODUCTION
Recently, as various mobile broadband equipments –
smart-phones, laptop PCs, tablet PCs, smart TVs and
more– have been widely diffused, global mobile
broadband traffic has been rapidly increasing (Cisco,
2011), and network traffic jams have resulted. Thus,
to maintain the service quality of mobile
communications, accurate forecasting results for
future mobile broadband traffic are required. When
the dominant traffic of mobile broadband networks
was voice, the number of subscribers of mobile
communication services affected patterns of traffic.
Therefore, forecasting results for mobile service
subscribers using diffusion models such as Bass
model (Bass, 1969), Logistic model (Mansfield,
1968) can be utilized as an alternative way for
mobile broadband traffic forecasting (Chu et al.,
2009). However, since new mobile broadband
services–web browsing, video streaming, online
game and more –which can be served by various
equipments have emerged and service usage patterns
have been diversified, it is hard to derive accurate
forecasting results using existing forecasting
methods. The need to reflect these changes to
forecasting has arisen. Therefore, this research aims
to suggest a new traffic forecasting method targeted
at the Korean mobile broadband market, considering
the number of users and patterns of use for mobile
broadband equipment.
The suggested forecasting method is based on a
three-round Delphi survey, and relevant actual data
was used to complement the survey result. The first
Delphi survey was conducted to identify mobile
broadband equipment and applications which are
expected to affect mobile broadband traffic in the
future. The second and third Delphi survey were
conducted to acquire further information on the
expected increase in of users and usage of
applications for each kind of equipment. Mobile
broadband traffic in the future was forecasted using
various demand forecasting methods that are
suitable for the characteristics of each kind of
equipment. The research results are expected to
provide the basic data for futher studies.
The remainder of this paper is organized as
follows. Section 2 reviews existing studies on the
demand diffusion model and Delphi survey. The
overall research framework and detailed processes are
presented in Section 3. Section 4 presents mobile
broadband traffic forecasting results. Finally, Section
5 discusses the summary, contributions, limitations,
and suggestions for further study of this research.
41
Cho C. and Lee S..
Mobile Broadband Traffic Forecasts in Korea.
DOI: 10.5220/0004066200410045
In Proceedings of the International Conference on Data Communication Networking, e-Business and Optical Communication Systems (DCNET-2012),
pages 41-45
ISBN: 978-989-8565-23-5
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
2 LITERATURE REVIEW
2.1 Demand Diffusion Models
The purpose of a diffusion model is to exhibit
aspects of innovation spreading among adopters in
the specific groups (Mahajan and Muller, 1979).
Since Bass (1969) suggested the concept, diffusion
models have been widely used for demand
forecasting. Diffusion models usually reflect internal
influence and external influence to forecasting. The
logistic model (Mansfield, 1968) and Gompertz
model (Steel, 1977) reflect internal influences by
imitations such as word-of-mouth to forecasting.
The bass model (Bass, 1969) reflects internal and
external influences (e.g. advertising effect)
simultaneously. Diffusion of mobile broadband
equipment is affected by internal and external
influences. Therefore, for the purpose of this study,
various demand diffusion forecasting models
considering mobile equipment’ characteristics were
used in forecasting.
2.2 Delphi Survey
A Delphi survey is developed as a systematic,
interactive forecasting method which relies on a
panel of experts. It usually consists of several rounds
to build a consensus on unclear things through
feedback processes. Thus, it has been used to derive
backup data in emerging industries which do not
have enough data for quantitative forecasting
(Bengisu and Nekhili, 2006; Gerdsri, 2003), or to
forecast the future of industries that need experts’
opinion to support quantitative forecasting results
such as the tourism industry (Yong et al., 1989) and
aerospace industry (English and Kernan, 1976).
Because mobile broadband equipment such as laptop
PCs, tablet PCs, and smart TVs have a lack of
relevant data, it is hard to apply existing forecasting
methods to them to predict demand or traffic.
Furthermore, since they are affected by external
factors (e.g. relevant policy) more easily than other
kinds of equipment, this fact needs to be considered.
Thus, this study utilized a Delphi survey to derive
backup data for equipment types which have a lack
of actual data.
3 RESEARCH FRAMEWORK
3.1 Overall Research Processes
The overall research processes of this study are as
follows (See figure 1.). First of all, the first Delphi
survey was conducted to identify devices (e.g. smart-
phones, tablet PCs, smart TVs) and applications (e.g.
voice, video streaming) that would affect the future
traffic needs. Then for the target devices, the
second- and third-round Delphi were carried out to
gather further information. Finally, based on the
results of the second- and third- Delphi survey, the
total amount of mobile broadband traffic in the
future was derived by summing up expected traffic
for each target device.
Figure 1: Overall research processes.
3.2 The First-round Delphi Survey
The first Delphi survey investigated the opinion of
18 experts in mobile communication industry. We
posed open questions to identify devices and
applications of each device. As a result, five mobile
broadband devices and three primary applications
for each device were identified (See Table 1.).
Table 1: Identified devices and applications.
Devices Applications
Feature phone Voice, Online game, VoIP
Smart phone
Online game, Web browsing, Video
streaming
Laptop PC
Web browsing, Video streaming,
Download
Tablet PC
Web browsing, Video streaming,
Download
Smart TV
Online game, Web browsing, Video
streaming
DCNET2012-InternationalConferenceonDataCommunicationNetworking
42
3.3 The Second- and Third- Delphi
Surveys
The second- and third- Delphi survey was conducted
to gather further information about identified
devices and applications in the first Delphi survey,
and 47 experts were involved to both. The second
Delphi survey was an attempt to determine –
information about the expected increase in rate of
users for each device and their patterns of use in
terms of “the type of applications frequently used in
the devices”. In the case of smart TV, due to its
relatively static usage pattern, the expected increase
in the rate of applications reflected the expected
increase in the rate of the number of smart TVs. The
third Delphi survey sought to build a consensus by
asking each participant identical questions, and the
survey results were used to forecast the total amount
of mobile broadband traffic in the future.
3.4 Mobile Broadband Traffic Forecast
The future mobile broadband traffic was forecasted
using the expected increase in the rate of users for
each device and the expected increased rate of usage
for each application. Traffic increases effects by an
increase of users and application usage in this way.
At first, the number of users for each device was
estimated according to their characteristics. Forecast
for the number of users for mobile phones–feature-
phones and smart-phones –was conducted using
actual data. For the other three devices, the expected
increase in the rate of users and the estimated initial
number of users were used. Next, a forecast for the
patterns of use for each device utilized the expected
rate of increase and weight of usage for each
application. Traffic consumptions for each
application were estimated using required
transmission speed ITU-R (International
Telecommunication Union-Radio communication
sector) definition. Last, the total amount of mobile
broadband traffic was calculated by reflecting
forecasts for the number of users and patterns of use.
Mobile broadband traffic was influenced by changes
in the number of users and usage patterns of
applications for each device. Thus, this study
developed a formula for the rate of increase in total
mobile broadband traffic which reflects the rate of
traffic increase by the number of users increase and
usage patterns of applications change (See formula
(1)).
T = U * A
the rate of increase in total mobile traffic (T)
the rate of increase in the number of users (U)
the rate of increase by usage pattern changing (A)
(1)
4 TRAFFIC FORECASTING
4.1 The Number of Users
The number of users for each device was forecasted
by diverse methods suitable for each in terms of
their characteristics. First of all, demand diffusion
models were used for mobile phones because
enough data for a quantitative forecast was available.
Among various demand diffusion models, those
which could provide powerful prediction and
interpretation were selected. The Bass model was
selected for feature-phones, and the Gompertz model
was chosen for smart-phones. Forecasts for the other
three devices were carried out by a panel of experts.
This method has been used in cases when relevant
data was not fulfilled, precise forecasting results
using data-based methods can not be provided, or
influences of external factors such as policy and
relevant law are extremely. Laptop PC usually has
functions which enable use of WiFi and Wireless
LAN networks that are easily affected by changes of
network policy. Tablet PCs and smart TVs have a
lack of data because they have not been widespread.
Therefore, a panel of experts attempted to forecast
the traffic for these three devices.
All forecasts were developed for five years from
2011 to 2016. Table 2 shows the forecasting results for
the increasing rate of users, and Figure 2 represents the
expected number of users for each device. While the
number of feature-phone users was expected to
decrease continuously, the number of the other four
devices’ users was expected to increase gradually.
Table 2: Forecasting results: the increasing rate of users.
Devices 2011 2012 2013 2014 2015 2016
Feature-
phone
1.00 0.69 0.70 0.82 0.93 0.99
Smart-
phone
1.00 1.91 1.32 1.12 1.05 1.02
Laptop 1.00 1.03 1.03 1.03 1.03 1.03
Tablets 1.00 1.15 1.15 1.15 1.15 1.15
Smart-TV 1.00 1.21 1.21 1.21 1.21 1.21
Figure 2: Forecasting results: the expected number of
users for each device.
MobileBroadbandTrafficForecastsinKorea
43
4.2 The Patterns of Use
The forecasting of the patterns of use for each device
was developed using the expected increased rate and
weight of use for each application. The initial
weights of use for each application were assumed to
be equal. Table 3 shows the forecasting results for
the increasing rate of application usages. For
feature-phones, use of VoIP was expected to grow
faster than others, as was while video streaming for
smart-phones. Video streaming and download were
expected to be the primary applications for laptop
PCs in the future, as are web browsing and video
streaming for tablet PC.
Table 3: Forecasting results: the expected increased rate of
application usages.
Applications
Voice
Online game
VoIP
Web browsing
Video
Streaming
Download
Devices
Feature-
phone
1.01 1.05 1.10 - - -
Smart-phone - 20 - 1.25 1.30 -
Laptop PC - - - 1.10 1.15 1.15
Tablet PC - - - 1.20 1.20 1.15
With the expected increased rate of application
usages, estimated traffic consumption of each
application was used to forecast the growth rates of
traffic due to the changes of usage patterns in
applications from 2011 to 2016. Traffic consumption
of each application is shown in Table 4. Among six
applications, voice and VoIP need the smallest level
of traffic, and online games, web browsing, and
downloads consume the middle level of traffic. The
largest amount of traffic is required for video
streaming.
Table 4: Estimated traffic consumption of each application
(ITU-R definition, kbps/second).
Applications Traffic consumption
VoIP 14
Voice 16
Online game 384
Web browsing 384
Download 384
Video streaming 2000
The expected growth rates of traffic due to the
changes of usage patterns in applications from 2011
to 2016 are as follows (See Table 5.). The expected
growth rates of feature-phones, laptops, and tablet
PCs were expected to be fixed while that of smart-
phones was expected to change dynamically. The
growth rate of traffic due to smart TV was replaced
by the growth rate of smart TV sales. It was
expected to grow continuously.
Table 5: Forecasting results: the growth rates of traffic due
to the changes of usage patterns in applications.
Devices 2011 2012 2013 2014 2015 2016
Feature-
phone
1.00 1.05 1.05 1.05 1.05 1.05
Smart-
phone
1.00 1.48 1.29 1.10 1.65 1.17
Laptop 1.00 1.14 1.14 1.14 1.14 1.14
Tablet 1.00 1.19 1.19 1.19 1.19 1.19
Smart
TV
1.00 1.30 1.30 1.30 1.10 1.10
4.3 Traffic Forecasting Results
To calculate the total amount of mobile broadband
traffic with formula (1), the increasing rate of users
in Table 2 and the growth rates of traffic due to the
changes of usage patterns in applications in Table 5
were used. Actual data were used as the initial traffic
for feature-phones and smart-phones. Others used
estimated traffic data. Table 6 describes how the
initial traffic data was acquired or estimated.
Table 6: The initial traffic of each device.
Devices Description
Featurephone Actual data
Smartphone Actual data
Laptop PC
estimated traffic per laptop * the
number of laptop * the number of
service subscribers
Tablet PC
estimated traffic per tablet * the
number of tablet * the number of
service subscribers
Smart TV
estimated traffic per smart TV * sales
of smart TV
Mobile broadband traffic including the growth of
users and changes of application usages by devices
and total traffic is represented in Table 7. The
forecasting results have two remarkable
characteristics. First, mobile broadband traffic in the
future is expected to be centralized to specific
devices. In 2011, smart-phones caused 53.4% of
mobile broadband traffic, followed by smart TVs
(35.4%), tablet PCs (7.4%), feature-phones (1.9%)
and laptop PCs (1.7%). But in 2016, top two devices,
smart-phones and smart TVs, are expected to cause
more than 95% of total mobile broadband traffic
(See Figure 3.). This means that mobile broadband
DCNET2012-InternationalConferenceonDataCommunicationNetworking
44
traffic forecasting targeted to specific devices can
derive feasible results.
Table 7: Total traffic increases and forecasted traffic of
devices.
Devices 2011 2012 2013 2014 2015 2016
Feature
- phone
T 0.34 0.24 0.18 0.16 0.15 0.16
R 1.00 0.72 0.74 0.86 0.98 1.04
Smart-
phone
T 9.79 27.6 47.1 58.1 100.1 119.6
R 1.00 2.82 1.71 1.23 1.72 1.19
Laptop
PC
T 0.31 0.37 0.43 0.51 0.60 0.71
R 1.00 1.18 1.18 1.18 1.18 1.18
Table
t PC
T 1.35 1.85 2.53 3.46 4.74 5.91
R 1.00 1.37 1.37 1.37 1.37 1.37
Smart
TV
T 6.47 9.31 13.4 19.3 27.7 36.4
R 1.00 1.44 1.44 1.44 1.44 1.31
Total
T
18.3 39.4 63.7 81.5 133 163
R
1.00 2.16 1.62 1.28 1.64 1.22
* T: The amount of expected traffic (1000TB)
R: The expected increasing rate of traffic
Second, the expected mobile broadband traffic
for each device was affected by changes in patterns
of use. In the case of smart-phones, although the
growth rate of users was expected to decrease
continuously, the growth rate of traffic was expected
to increase similarly with the growth rate of traffic
due to the changes of usage patterns. This trend also
appeared in the case of smart TVs. Thus, patterns of
usage and their changes should be considered among
the most important aspects for mobile broadband
traffic forecasting.
5 CONCLUSIONS
This study aims to suggest a new traffic forecasting
method considering the number of users and patterns
of use for mobile broadband equipment. For the
purpose of this study, a three-round Delphi survey
was conducted to identify devices and applications
that would affect the future traffic needs. Then,
forecasts of the number of users for devices and the
growth rates of traffic caused by the changes in
usage patterns of applications were accomplished.
As a result, smart-phones and smart TVs were
identified as requiring most of the traffic needs in
the future, and the changes of usage patterns were
expected to influence the total amount of mobile
broadband traffic considerably. There are two main
contributions of this study. First, this study suggests
a forecasting method which reflects the changes of
usage patterns for mobile broadband devices to
support deriving realistic results. Next, this study
can provide the basic research data for future studies
in the mobile communication area and forecasting in
other areas.
However, this study has also a limitation. In the
suggested forecasting method, there exist several
assumptions. This is inevitable because it was
caused by the characteristics of the devices. If actual
data is used in the forecasting, better results can be
acquired. Therefore, in a further study, this will be
complemented to derive more reliable forecasting
results.
ACKNOWLEDGEMENTS
This research was supported by Basic Science
Research Program through the National Research
Foundation of Korea (NRF) funded by the Ministry
of Education, Science and Technology (No. 2009-
0089674).
REFERENCES
Bass, F. M., 1969. A New Product Growth Model for
Consumer Durables. Management Science, Vol. 15,
No. 5, pp. 215-227.
Bengisu, M., Nekhili, R., 2006. Forecasting emerging
technologies with the aid of science and technology
databases. Technological Forecasting and Social
Change, Vol. 73, No. 7, pp. 835-844.
Chu, W., Wu, F., Kao, K., Yen, D., 2009. Diffusion of
mobile telephony: An empirical study in Taiwan.
Telecommunications Policy, Vol. 33, No. 9, pp. 506-
520.
CISCO, 2011. Cisco Visual Networking Index: Global
Mobile Data Traffic Forecast Updata, 2010-2015.
white paper.
English, J. M., Kernan, G. L., 1976. The prediction of air
travel and aircraft technology to the year 2000 using
the Delphi method. Transportation Research, Vol. 10,
No. 1, pp. 1-8.
Gerdsri, N., 2003. An analytical approach to building a
technology development envelope (TDE) for
roadmapping of emerging technologies: a case study
of emerging electronic cooling technologies for
computer servers. In the Proceedings of PICMET’03,
pp. 380-389.
Mahajan, V., Muller, E., 1979. Innovation Diffusion and
New Product Growth Models in Marketing. Journal of
Marketing, Vol. 43, No. 4, pp. 55-68.
Mansfield, E., 1968. Industrial research and technological
innovation: An econometric analysis. New York:
Norton.
Steel, G. G., 1977. Growth Kinetics of Tumors. Oxford:
Clarendon Press.
MobileBroadbandTrafficForecastsinKorea
45