future internet

Article
Short-Term Mobile Network Traffic Forecasting Using Seasonal
ARIMA and Holt-Winters Models
Irina Kochetkova 1,2, * , Anna Kushchazli 1 , Sofia Burtseva 1 and Andrey Gorshenin 2, *

1 Institute of Computer Science and Telecommunications, RUDN University, 6 Miklukho-Maklaya St.,

117198 Moscow, Russia; aikushch@yandex.ru (A.K.); sofiya_burceva@inbox.ru (S.B.)
2 Institute of Informatics Problems, Federal Research Center “Computer Science and Control” of the Russian
Academy of Sciences, 44-2 Vavilova St., 119333 Moscow, Russia
* Correspondence: kochetkova-ia@rudn.ru (I.K.); agorshenin@frccsc.ru (A.G.)

Abstract: Fifth-generation (5G) networks require efficient radio resource management (RRM) which
should dynamically adapt to the current network load and user needs. Monitoring and forecasting
network performance requirements and metrics helps with this task. One of the parameters that
highly influences radio resource management is the profile of user traffic generated by various 5G
applications. Forecasting such mobile network profiles helps with numerous RRM tasks such as
network slicing and load balancing. In this paper, we analyze a dataset from a mobile network
operator in Portugal that contains information about volumes of traffic in download and upload
directions in one-hour time slots. We apply two statistical models for forecasting download and
upload traffic profiles, namely, seasonal autoregressive integrated moving average (SARIMA) and
Holt-Winters models. We demonstrate that both models are suitable for forecasting mobile network
traffic. Nevertheless, the SARIMA model is more appropriate for download traffic (e.g., MAPE [mean
absolute percentage error] of 11.2% vs. 15% for Holt-Winters), while the Holt-Winters model is better
suited for upload traffic (e.g., MAPE of 4.17% vs. 9.9% for SARIMA and Holt-Winters, respectively).

Keywords: 5G; mobile network traffic; download; upload; forecasting; time series; ARIMA; SARIMA;
Citation: Kochetkova, I.; Kushchazli,
Holt-Winters
A.; Burtseva, S.; Gorshenin, A.
Short-Term Mobile Network Traffic
Forecasting Using Seasonal ARIMA
and Holt-Winters Models. Future
1. Introduction
Internet 2023, 15, 290. https:// Fifth-generation (5G) and 6G networks [1,2] are expected to support a wide range of
doi.org/10.3390/fi15090290 new technologies, such as drones and virtual/augmented reality, which require high bit
Academic Editors: Ammar
rates, lower latency, and increased throughput [3–5]. The number of connected devices
Muthanna and Mohammed
is increasing, resulting in a dramatic growth in traffic volume, causing anomalies such as
Abo-Zahhad network congestion, decreased quality of service, network delays, data loss, and blocking
of new connections [6]. The network architecture should adapt to the volumes of traffic
Received: 3 August 2023 generated by various applications and use it for decision-making, taking into account
Revised: 22 August 2023 several types of traffic with different service and priority requirements [7–9]. Artificial
Accepted: 26 August 2023
intelligence (AI) and machine learning (ML) are now trends for 5G networks that could
Published: 28 August 2023
provide more efficient and reasonable network planning and management [10]. AI and ML
models could be trained on a large amount of data that service providers collect [11]. The
collected data should be reliable, and the analysis carried out should be accurate [12].
Copyright: © 2023 by the authors.
Network traffic forecasting is one of the tasks that use ML methods for effective net-
Licensee MDPI, Basel, Switzerland.
work management [13,14]. This task aims to identify potential problems before they occur,
This article is an open access article reduce service outages, manage user needs, and analyze user behavior in applications [15].
distributed under the terms and For example, traffic forecasting is used for smart power consumption by a base station [11].
conditions of the Creative Commons In [16], this problem is considered based on network slicing, mobile edge computing, base
Attribution (CC BY) license (https:// station sleeping, and additional power during high-demand hours. Traffic forecasting is
creativecommons.org/licenses/by/ divided into short-term and long-term, but sometimes medium-term forecasting is also
4.0/). necessary [17].

Future Internet 2023, 15, 290. https://doi.org/10.3390/fi15090290 https://www.mdpi.com/journal/futureinternet

Future Internet 2023, 15, 290 2 of 15

1.1. Related Work

Forecasting methods typically begin with statistical models, such as ARIMA (autore-
gressive integrated moving average) [18] and its seasonal variant SARIMA, exponential
smoothing (Holt-Winters, extended Holt-Winters, etc.), and regression models (multiple,
linear, etc.). In addition, ML and deep learning models, Gaussian models (Gaussian mixture
model [19,20], Gaussian process), random forests, and neural networks are also commonly
used. While ML models generally outperform statistical models, there are instances where
statistical models are applied for their faster processing capabilities. To achieve more
accurate results, a combination of different approaches is often utilized [16]. The same
applies to network traffic forecasting. LTE (long-term evolution) traffic is predicted by the
ARIMA model [21], as well as bagging, random forest, and support vector machines [22].
In [23], the authors used ARIMA for post-processing the residuals of the ML algorithm,
which improves the accuracy of traffic prediction. In [24], 5G traffic is forecasted by gated
recurrent unit and long short-term memory (LSTM) networks. The authors of [25,26] show
that LSTM is good for online forecasting.
Let us review selected papers on the subject of applying statistical models in network
traffic forecasting. Table 1 provides a summary of these papers in terms of the tasks they
address, the applications that generate traffic, the models used, and the metrics used to
evaluate these models. Most of the papers focus on traffic forecasting, but the tasks of
capacity planning [27] and resource optimization [28] are also addressed. The data are
collected from several sources, including cells [29–32], devices [33], switches [34], and
servers [35]. The most commonly used statistical model is the ARIMA model [36,37].
Moreover, combinations of statistical and machine learning methods are used to achieve
more accurate results [38,39]. The evaluation metrics help to determine which solutions are
most suitable for the proposed model. The choice of metrics depends on the specific study,
such as MAE and MSE [30,31], or performance indicators [40].

Table 1. Summary of selected works on statistical models for network traffic forecasting.

Ref. Task Data Source/Application Model Evaluation Metric

[30] Traffic forecasting Traffic from 2 cells ARIMA MAE, NMSE
[31] Traffic forecasting Traffic from 3 LTE cells ARIMA, LSTM MSE, MAE, R2 score
Random walk, Linear trend,
Circuit switch and packet
[34] Network capacity Exponential smoothing, RMSE, MAPE
switch 3G traffic
forecasting ARIMA
SARIMA, Holt-Winters, Ran-
[38] Traffic forecasting Traffic from 7160 LTE cells MAPE, MAE
dom Forest, SVM, ANN
Traffic from the network
[28] Recourse optimiza- with discontinuous reception ARIMA, LSTM RMSE
tion (DRX) scheme
Random walk, Historical
Traffic from 8 detectors of vir-
[39] Traffic congestion mean, ARIMA, LSTM, GRU, RMSE, MAE, MAPE
tual reality railway
forecasting DCFCN
Traffic from 470 access points Holt-Winters, SARIMA, MAE, RMSE, NRMSE, R2
[36] Traffic forecasting
of an enterprise network LSTM, GRU, CNN score
Efficiency, switches per
Download traffic from HSPA
[40] Throughput forecast- ARIMA, FARIMA, ANN minute and buffering per
network
ing minute
WiFi and cellular download
[35] Traffic forecasting ARIMA, FARIMA, SVR, RNN MAE, RMSE, MAPE, MASE
traffic from a server
Error rate, average absolute
[33] Traffic forecasting WiFi traffic from 15 protocols ARIMA deviation, mean and variance
of the error
Future Internet 2023, 15, 290 3 of 15

Table 1. Cont.

Ref. Task Data Source/Application Model Evaluation Metric

Traffic from wireless sensor Complexity, Accuracy, Intelli-
[41] Traffic forecasting ARIMA
network gence, Independence
and anomaly detec-
tion
[32] Traffic forecasting Traffic from 191 eNodeB ARIMA AIC, AICc, BIC
ARIMA, FARIMA, SVM,
[42] Traffic forecasting WLAN traffic MSE, NMSE
Welevet, ANN
Voice and data traffic from 600 Exponential Smoothing, Holt-
[29] Traffic forecasting AIC, SSR, RMSE, AMSE
cells Winters
MAE, MSE, RMSE, NRMSE,
[37] Spectrum efficiency Traffic from 3 countries AR, MA, ARMA, ARIMA
NMAE
forecasting
Traffic from an institutional
[43] Traffic forecasting ARIMA, ANFIS RMSE
wireless network
Percentage error between
[44] Traffic forecasting LTE and 3G traffic ARIMA
models

1.2. Contributions
In this paper, we analyze mobile network traffic collected from a network in Portugal
over a half-month period. The available data include the number of megabytes (MB) sent
and received by various applications during each hour. Our goal is to forecast the total
traffic behavior separately for downlink and uplink using fast processing statistical models.
The main contributions of our study are as follows:
• We analyze the dataset of real network traffic from a mobile operator in Portugal using
fast processing statistical models, namely SARIMA, and Holt-Winters, which have not
been applied to this data before.
• We demonstrate that the SARIMA model is more appropriate for forecasting download
traffic, while the Holt-Winters model is better suited for forecasting upload traffic,
showing appropriate errors in the considered dataset.
• Since statistical models are suitable for fast and precise forecasting of mobile network
traffic, they can be implemented in cellular operators’ solutions without a significant
increase in cost.
The rest of the paper is organized as follows. Section 2 provides a description of the
dataset and illustrates the traffic behavior, both in total and by various applications. In
Section 3, we discuss the SARIMA and Holt-Winters models, along with the necessary pre-
liminary checks. Section 4 outlines the metrics used for evaluating the models and presents
forecasts for both download and upload traffic. Conclusions are drawn in Section 5.

2. Descriptive Statistics
In this section, we will describe the dataset and analyze the overall traffic profile, in-
cluding the total traffic over all applications and the average traffic for various applications.

2.1. Dataset Description

We use the dataset obtained from a mobile network operator that offers multiple
services (applications), such as Internet access, messaging, calls, file transfer, etc. The data
flow is bidirectional, with traffic flowing from the base station to the user device in the
downlink direction and from the device to the base station in the uplink direction. Each
user device is associated with a unique masked mobile station international subscriber
directory number (MSISDN). The monitoring system records the upload and download
data generated by each device for each application class every hour. The statistics represent
the volume of traffic in megabytes (MB) for both the download and upload directions.
Variable descriptions are provided in Table 2, and dataset records are presented in
Table 3. The dataset comprises 41,479,488 records containing information for 15 days, with
Future Internet 2023, 15, 290 4 of 15

a total of 94,632 users’ devices. Table 4 displays the list of application classes along with the
corresponding number of records. There are 16 application classes with the largest being
“Web Applications”, which is a client-server application that allows a user to interact with a
web server using a browser. The top three most frequently used applications by users are
also “Instant Messages Applications” and online “Games”. The least used classes include
“Legacy Protocols” and “DB Transactions”, which are responsible for working with database
systems and file systems. The“Others” category represents non-classified applications not
related to user traffic. These may include technical services such as directory services,
network management services, automatic network address configuration, and mapping for
location determination.

Table 2. Dataset variables.

Variable Description
START HOUR Start time of the one-hour period for measuring traffic
MASKED MSISDN Masked mobile station international subscriber directory number
APP CLASS Application class
UPLOAD Incoming traffic in the uplink during one hour [MB]
DOWNLOAD Outgoing traffic in the downlink during one hour [MB]

Table 3. Dataset records example.

START HOUR MASKED MSISDN APP CLASS UPLOAD DOWNLOAD

2018-02-10 01:00:00 F6C1745A0A9DF638DE2C14683E0F250D Streaming Applications 0.002527 0.000616
2018-02-10 01:00:00 6474B3E3E20B5887A7593C61439250A9 Others 0.000828 0.000334
2018-02-10 01:00:00 B05DEBB3D0E2ACD68FE47611CA3FDDCB Web Applications 0.000967 0.001813
2018-02-10 01:00:00 B2D5516431ECC5B6851FD9FBAE0387A7 Games 0.000039 0.000052
2018-02-10 01:00:00 5B507EECA75149121F9C86E8690109D8 Mail 0.012802 0.006137

Table 4. Application classes.

Application No. Records Application No. Records

Web Applications 9,641,283 Others 6,464,018
Instant Messaging Applications 5,540,289 Games 5,199,684
File Transfer 4,259,626 Mail 2,622,758
Streaming Applications 2,420,552 VoIP 1,825,777
Security 1,667,752 Music Streaming 790,107
Network Operation 635,634 P2P Applications 274,245
Terminals 121,807 File Systems 10,118
DB Transactions 5816 Legacy Protocols 22

2.2. Total Traffic Behavior

Let us examine traffic behavior in this study. Table 5 provides further insight with
descriptive statistics. For a specific user and application, the traffic can only flow in one
direction. Therefore, there cannot be any upload traffic with non-zero download and vice
versa, resulting in the minimum values of 0 in Table 5.

Table 5. Descriptive statistics for total traffic.

Upload Download
Mean 0.5442475 0.05518938
Standard deviation 5.322763 1.702829
Minimum 0 0
Mode 0.000029 0.00001
Maximum 2324.251 1337.792
Future Internet 2023, 15, 290 5 of 15

Specifically, let us consider upload traffic. The standard deviation is 5.322763, and the
mean is 0.5442475. These values suggest that the ratio of the standard deviation to the array
values of the samples differs, indicating that the values are distributed over a wider range
of data values. For download traffic, the situation is comparable: the standard deviation
is 1.702829, and the mean is 0.05518938, with a significant difference between the array
values.
To assess the relationship between upload and download traffic, we used Spearman’s
rank correlation coefficient, or Spearman’s ρ [18]. We chose this non-parametric method
instead of the parametric Pearson method due to the fact that the Pearson criterion is
applied to two quantitative indicators that have a linear relationship. Spearman’s method
can be applied to any set of data without requiring additional preparation and processing
of the values. Essentially, it allows for the determination of the strength of the relationship.
For upload and download traffic with statistically significant differences at ρ < 0.01, a
correlation coefficient of 0.914 indicates a very high correlation strength between upload
and download traffic.

2.3. Traffic by Applications

Let us divide the application classes into three groups, as shown in Table 6. Figures 1–3
illustrate the traffic profile in each group. Group No. 1 (Figure 1) is similar to the total
traffic profile, and the time series is seasonal. For example, for “Web applications”, the
correlation coefficient for upload and download traffic is 0.961. Group No. 2 (Figure 2)
is not similar to the total traffic profile and is non-seasonal with outliers. In “Terminals”
application, the correlation coefficient is 0.935. Group No. 3 (Figure 3) is not similar to the
total traffic profile and is seasonal. For example, in “VoIP”, the coefficient is 0.708.

1.6
Upload
Download
1.4

1.2

1.0
Values, MB

0.8

0.6

0.4

0.2

0.0
11 12 13 14 15 16 17 18 19 20 21 22 23 24
Feb
2018
Date
Figure 1. “Web Applications” traffic (group of applications No. 1).

2.5
Upload
Download

2.0

1.5
Values, MB

1.0

0.5

0.0
11 12 13 14 15 16 17 18 19 20 21 22 23 24
Feb
2018
Date
Figure 2. “Terminals” traffic (group of applications No. 2).
Future Internet 2023, 15, 290 6 of 15

0.40 Upload
Download
0.35

0.30

0.25

Values, MB
0.20

0.15

0.10

0.05

0.00
11 12 13 14 15 16 17 18 19 20 21 22 23 24
Feb
2018
Date
Figure 3. “VoIP” traffic (group of applications No. 3).

Table 6. Groups of applications.

Group Applications
1. Time series is similar to the total traffic profile Others, Streaming Applications, Web Applica-
and seasonal tions
DB Transactions, File Systems, File Transfer,
2. Time series is not similar to the total traffic
Games, Mail, Music Streaming, P2P Applica-
profile and non-seasonal with outliers
tions, Security, Terminals
3. Time series is not similar to the total traffic Instant Messaging Applications, Legacy Proto-
profile and seasonal cols, VoIP, Network Operation

3. Statistical Models for Forecasting Traffic

In this section, we provide a description and formulas for two models that we will
use for forecasting mobile network traffic: the SARIMA and Holt-Winters models. Table 7
includes the notations used for the parameters of the SARIMA and Holt-Winters models.
Table 7. Main notation.

Parameter Description
Time series parameters
xt Time series of data
yt Forecast value of xt
SARIMA model parameters
p Order of the non-seasonal AR part
d Degree of differencing for the non-seasonal part
q Order of the non-seasonal MA part
P Order of the seasonal AR part
D Degree of differencing for the seasonal part
Q Order of the seasonal MA part
m Number of periods in each season
Lxt = xt−1 , Li xt = xt−i Lag operator
εt Error terms
φi Parameters of the non-seasonal AR part
θi Parameters of the non-seasonal MA part
Φi Parameters of the seasonal AR part
Θi Parameters of the seasonal MA part
Holt-Winters model parameters
st Smoothed value of xt
bt Estimate of the trend
ct Seasonal change factor
α Data smoothing factor
β Trend smoothing factor
γ Seasonal change smoothing factor
Future Internet 2023, 15, 290 7 of 15

3.1. Seasonal ARIMA Model

The seasonal autoregressive integrated moving average (SARIMA) model is an ex-
tension of the ARIMA model that explicitly supports univariate time series data with a
seasonal component [45–47]. The model comprises three hyperparameters that define the
autoregressive (AR), integrated (I), and moving average (MA) for the non-seasonal com-
ponent of the time series, as well as an additional hyperparameter for the seasonal period
(S). The model is denoted as SARIMA( p, d, q)( P, D, Q)m , where the first three parameters
( p, d, q) refer to the non-seasonal part of the model (ARIMA), whereas the last parameters
( P, D, Q)m represent the seasonal part. Specifically, p is the order (number of time lags) of
the non-seasonal AR part, d is the degree of differencing (the number of times the data have
had past values subtracted) for the non-seasonal part, q is the order of the non-seasonal MA
part, P is the order of the seasonal AR part, D is the degree of differencing for the seasonal
part, Q is the order of the seasonal MA part, and m is the number of periods in each season.
Given a time series of data xi , i = 0, . . . , t − 1, in order to calculate the forecast value
of yt = xt at time t, the SARIMA model is used and written as

p
! !
P
1 − ∑ φi L i
1 − ∑ Φi L im
(1 − L ) d (1 − L m ) D x t
i =1 i =1
q
! !
Q
= 1 + ∑ θi Li 1 + ∑ Θi Lim ε t ,
i =1 i =1

where Lxt = xt−1 and Li xt = xt−i represent the lag operator, ε t are the error terms, φi
are the parameters of the non-seasonal AR part of the model, θi are the parameters of the
non-seasonal MA part, Φi are the parameters of the seasonal AR part, Θi are the parameters
of the seasonal MA part.

3.2. Holt-Winters Model

The Holt-Winters model, also known as triple exponential smoothing, is used to
predict time series data that exhibit both trend and seasonal variations [18,46,47]. There are
different types of trends and seasonality: additive and multiplicative in nature, meaning
linear and exponential, respectively.
Given a time series of data xi , i = 0, . . . , t − 1, let us denote the forecast value of xt
at time t as yt , the smoothed value of xt at time t as st , the estimate of the trend at time t
as bt , and the seasonal change factor at time t as ct . Depending on the types of trend and
seasonality, the following formulas represent the Holt-Winters model [48]:
• additive (linear) trend and additive (linear) seasonality

yt+h = st + hbt + ct+h−m(b h−1 c+1) ,

m

st = α( xt − ct−m ) + (1 − α)(st−1 + bt−1 ),

bt = β ( s t − s t − 1 ) + ( 1 − β ) bt − 1 ,
c t = γ ( x t − s t ) + (1 − γ ) c t − m ;

• multiplicative (exponential) trend and additive (linear) seasonality

yt+h = st · hbt + ct+h−m(b h−1 c+1) ,

m

s t = α ( x t − c t − m ) + ( 1 − α ) s t − 1 · bt − 1 ,
st
bt = β + ( 1 − β ) bt − 1 ,
s t −1
c t = γ ( x t − s t ) + (1 − γ ) c t − m ;
Future Internet 2023, 15, 290 8 of 15

• additive (linear) trend and multiplicative (exponential) seasonality

yt+h = (st + hbt ) · ct+h−m(b h−1 c+1) ,

m
xt
st = α + (1 − α)(st−1 + bt−1 ),
ct−m
bt = β ( s t − s t − 1 ) + ( 1 − β ) bt − 1 ,
xt
c t = γ + (1 − γ ) c t − m ;
st

• multiplicative (exponential) trend and multiplicative (exponential) seasonality

yt+h = st · hbt · ct+h−m(b h−1 c+1) ,

m
xt
st = α + ( 1 − α ) s t − 1 · bt − 1 ,
ct−m
st
bt = β + ( 1 − β ) bt − 1 ,
s t −1
xt
c t = γ + (1 − γ ) c t − m ;
st

where α is the data smoothing factor, β is the trend smoothing factor, and γ is the seasonal
change smoothing factor.

3.3. Preliminary Checks

Before applying the SARIMA and Holt-Winters models, some preliminary checks
should be conducted [49]. Both models require that the data series exhibit seasonality. For
this purpose, the STL (seasonal and trend decomposition using Loess) decomposition [50]
can be employed. The seasonal period is 24 h and is associated with user activity. The
seasonal variation around each level appears to increase proportionally with the current
levels. Therefore, seasonality may be multiplicative. Regarding the trend, there are no
significant changes in the lines, and their slopes are close to zero. Therefore, we may
consider the trend as additive.
To verify the stationarity of the time series, we employed the ADF (Augmented Dickey–
Fuller) test [51] and the KPSS (Kwiatkowski–Phillips–Schmidt–Shin) test. In both tests, the
test statistic should be less than the significance level of α = 0.05. The ADF test assesses the
null hypothesis that the time series is not stationary. For both upload and download traffic,
the p-values are less than 0.05, with values of 0.00063973 and 0.000000725, respectively. The
null hypothesis for the KPSS test is opposite to that of the ADF test. The p-value for the
KPSS test is 0.1 in both cases, which exceeds α = 0.05. Therefore, we can conclude that the
time series is stationary.

4. Forecasting Download and Upload Traffic

In this section, we will begin with discussing the metrics used to evaluate the models,
and then move on to forecasting download and upload traffic.

4.1. Evaluation Metrics

The first step is to choose the parameters for the models. For the SARIMA model, we
choose the parameters p, d, q, P, D, Q using the Akaike information criterion (AIC) [46] in
order to minimize it. Specifically, we use the following equation for AIC:

AIC = 2k − 2 log( L) = 2( p + q + P + Q) − 2 log( L),

where k is the number of estimated parameters in the model, and L is the maximized value
of the likelihood function for the model. For the Holt-Winters model, the equations depend
on the types of trend and seasonality, namely, additive and multiplicative. We perform
brute force checks to determine the appropriate equations.
Future Internet 2023, 15, 290 9 of 15

The second step is to compare different models. We use typical evaluation metrics [52,53]
such as mean squared error (MSE), root mean square error (RMSE), mean absolute error
(MAE), mean absolute percentage error (MAPE), and mean squared logarithmic error
(MSLE). Table 8 provides a summary of these metrics. In our notation, xt represents the
time series data and yt denotes the forecasted value of xt .
The dataset was normalized and divided into a training dataset of 13 days and testing
datasets of 2 days, which is approximately 13%. We used Python for programming and its
modules statsmodels and pmdarima.

Table 8. Metrics for traffic forecast evaluation.

Metric Formula

1 n
n t∑
Mean squared error (MSE) ( x t − y t )2
=1

s
1 n
n t∑
Root mean square error (RMSE) ( x t − y t )2
=1

1 n
n t∑
Mean absolute error (MAE) | xt − yt |
=1

1 n xt − yt
n t∑
Mean absolute percentage error (MAPE) · 100%
=1 xt

1 n
n t∑
Mean squared logarithmic error (MSLE) [log( xt + 1) − log(yt + 1)]2
=1

4.2. Download Traffic

For download traffic, the parameters for the SARIMA model are as follows:

SARIMA(2, 0, 1)(0, 0, 2)24 .

For the Holt-Winters model, the trend appears to be multiplicative and the seasonality
additive.
Figures 4 and 5 show three-day forecast plots for download traffic using the SARIMA
and Holt-Winters models, respectively. The black line represents the training data, the black
dashed line represents the test data, the forecast is shown in purple, and the red line shows
the forecast for comparison with the actual data. Table 9 summarizes the evaluation metrics
that demonstrate the superiority of the Holt-Winters model in forecasting download traffic.

Table 9. Metrics for download traffic forecast evaluation.

Metric SARIMA Model Holt-Winters Model

MSE 0.0181 0.00021
RMSE 0.0181 0.0145
MAE 0.01513 0.01217
MAPE 15% 11.2%
MSLE 0.000258 0.000163
Future Internet 2023, 15, 290 10 of 15

SARIMA

0.08
Averague values for download, MB

0.06

0.04

0.02
Training data
Forecast for comparison
Test data
0.00 Predicted values
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Feb
2018
Figure 4. Download traffic forecastDate
using SARIMA model.

Holt-Winters
0.10 Training data
Test data
Forecast for comparison
Predicted values
0.08
Averague values for download, MB

0.06

0.04

0.02

11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Feb
2018
Figure 5. Download traffic forecastDate
using Holt-Winters model.

4.3. Upload Traffic

For upload traffic, the parameters for the SARIMA model are as follows:

SARIMA (3, 0, 1)(2, 0, 2)24

For the Holt-Winters model, the trend appears also to be multiplicative and the seasonality
additive.
The forecast for upload traffic is shown in Figures 6 and 7 using the SARIMA and Holt-
Winters models, respectively. By comparing the test and forecast data, we can conclude
that the dynamics of the forecast peaks for upload traffic do not repeat the test values.
Additionally, the fluctuations exhibit pronounced general seasonality. However, if we
consider day 24, the SARIMA model shows better results as it accurately predicts the peak
in the data. On the other hand, the Holt-Winters model assumes peaks in days after day 24,
resulting in fluctuations that have pronounced overall seasonality and are more similar to
the test data. Based on the results in Table 10, it can be concluded that the SARIMA model
is more accurate in predicting upload traffic in our case.
Future Internet 2023, 15, 290 11 of 15

SARIMA

0.8
Averague values for upload, MB

0.6

0.4

0.2
Training data
Forecast for comparison
Test data
0.0 Predicted values
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Feb
2018
Date SARIMA model.
Figure 6. Upload traffic forecast using

Holt-Winters
Training data
Test data
Forecast for comparison
0.8 Predicted values
Averague values for upload, MB

0.6

0.4

0.2

11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Feb
2018
Date Holt-Winters model.
Figure 7. Upload traffic forecast using

Table 10. Metrics for upload traffic forecast evaluation.

Metric SARIMA Model Holt-Winters Model

MSE 0.00004 0.00015
RMSE 0.006 0.0123
MAE 0.0046 0.0099
MAPE 4.17% 9.9%
MSLE 0.00003 0.000118

4.4. Discussion
This study aimed to examine the effectiveness of using SARIMA and Holt-Winters
models for short-term forecasting of mobile network traffic. The findings of the study
indicate that both models can yield valuable insights for predicting future traffic in mobile
networks. The SARIMA model has been recognized for its capability to capture temporal
patterns in time series data. It exhibited effectiveness in capturing short-term fluctuations
and trends in mobile network traffic. Additionally, the Holt-Winters seasonal model,
designed to account for the inherent seasonality in time series data, was also explored. By
incorporating seasonal components such as trend and seasonality, the Holt-Winters model
successfully captured cyclical patterns of mobile network traffic.
Future Internet 2023, 15, 290 12 of 15

To assess the forecast results, we computed various evaluation metrics, including MSE,
RMSE, MAE, MAPE, and MSLE. The results, presented in Tables 9 and 10, demonstrated
the suitability of each model for different datasets. In an effort to facilitate comparison
between the predicted and test data, distinct lines were plotted on Figure 8 with four lines
representing the absolute error indicators for both models and the traffic directions.

real upload - forecast by SARIMA

real upload - forecast by HW
0.20 real download - forecast by SARIMA
real download - forecast by HW
Absolute error (modulus)

0.15

0.10

0.05

0.00
00:00 06:00 12:00 18:00 00:00 06:00 12:00 18:00
23-Feb 24-Feb
2018
Figure 8. Absolute error for traffic Date
forecast.

5. Conclusions
The number of users and equipment is growing extremely quickly, and telecom
operators need to understand the demand for different types of applications in next-
generation networks. The ability to predict such demand would help service providers
make better offers to customers. This paper has explored the use of statistical methods of
data analysis in the context of 5G networks. The main objective of this study was to analyze
mobile network traffic and develop forecasting models for traffic profiles. Two statistical
models, SARIMA and Holt-Winters, were constructed and evaluated for this purpose. The
results demonstrate that both models effectively predict the average values of upload and
download traffic within a certain range. However, it was observed that the Holt-Winters
model is better suited for forecasting download traffic profiles, while SARIMA is more
suitable for upload traffic profiles.
From our numerical analysis, we found that each statistical method has its own
specifications. There is no universality, as each dataset requires its own approach. For
example, the MAPE for download traffic was 11.2% for SARIMA and 15% for Holt-Winters.
However, the Holt-Winters model was better suited for upload traffic, with a MAPE of
4.17% compared to 9.9% for SARIMA and Holt-Winters, respectively. Additionally, we
observed that the MSE metric for download traffic was 86 times less for the Holt-Winters
model (0.00021) compared to SARIMA (0.0181). Conversely, for upload traffic, the MSE
was almost four times less for SARIMA (0.00004) compared to Holt-Winters (0.0015).
Future studies will focus on combining statistical models with machine learning
methods for more precise forecasts, as well as anomaly detection. By implementing such
techniques, we aim to enhance the accuracy and reliability of traffic forecasting in 5G
networks. These findings contribute to the growing body of knowledge surrounding the
utilization of data analysis methods in the field of telecommunications.

Author Contributions: Conceptualization, project administration, supervision, methodology,

writing—review and editing, I.K. and A.G.; formal analysis, investigation, A.K. and I.K.; software,
validation, visualization, writing—original draft, A.K. and S.B. All authors have read and agreed to
the published version of the manuscript.
Future Internet 2023, 15, 290 13 of 15

Funding: This publication has been supported by the RUDN University Scientific Projects Grant
System, project No. 025319-2-000 (recipient I. Kochetkova). The research by A. Gorshenin has been
supported by the Ministry of Education and Science of the Russian Federation as part of the program
of the Moscow Center for Fundamental and Applied Mathematics under the agreement No. 075-
15-2022-284. The research was carried out using the infrastructure of the Shared Research Facilities
“High Performance Computing and Big Data” (CKP “Informatics”) of the Federal Research Center
“Computer Science and Control” of the Russian Academy of Sciences.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Acknowledgments: The authors are sincerely grateful to Luis M. Correia (IST/INESC-ID, University
of Lisbon) for providing the dataset.
Conflicts of Interest: The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript:

3GPP 3rd Generation Partnership Project

5G 5th generation
ADF Augmented Dickey–Fuller test
AI Artificial Intelligence
AIC Akaike information criterion
AR Auto regressive
ARIMA Auto regressive integrated moving average
ITU-T International Telecommunications Union – Telecommunications sector
KPSS Kwiatkowski–Phillips–Schmidt–Shin test
LSTM Long short-term memory
LTE Long-term evolution
MA Moving average
MAE Mean absolute error
MAPE Mean absolute percentage error
ML Machine learning
MSE Mean squared error
MSISDN Mobile station international subscriber directory number
MSLE Mean squared logarithmic error
P2P Peer-to-peer
RMSE Root mean squared error
SARIMA Seasonal ARIMA
VoIP Voice over internet protocol

References
1. Giordani, M.; Polese, M.; Mezzavilla, M.; Rangan, S.; Zorzi, M. Toward 6G Networks: Use Cases and Technologies. IEEE Commun.
Mag. 2020, 58, 55–61. [CrossRef]
2. Saad, W.; Bennis, M.; Chen, M. A Vision of 6G Wireless Systems: Applications, Trends, Technologies, and Open Research
Problems. IEEE Netw. 2020, 34, 134–142. [CrossRef]
3. Campos, R.; Ricardo, M.; Pouttu, A.; Correia, L. Wireless Technologies Towards 6G. Eurasip J. Wirel. Commun. Netw. 2023, 2023.
[CrossRef]
4. Kochetkov, D.; Vuković, D.; Sadekov, N.; Levkiv, H. Smart Cities and 5G Networks: An Emerging Technological Area? J. Geogr.
Inst. Jovan Cvijic SASA 2019, 69, 289–295. [CrossRef]
5. Kochetkov, D.; Almaganbetov, M. Using Patent Landscapes for Technology Benchmarking: A Case of 5G Networks. Adv. Syst.
Sci. Appl. 2021, 21, 20–28. [CrossRef]
6. Ruiz, S.; Ahmadi, H.; Gardašević, G.; Haddad, Y.; Katzis, K.; Grazioso, P.; Petrini, V.; Reichman, A.; Ozdemir, M.; Velez, F.; et al.
5G and Beyond Networks; Elsevier: Amsterdam, The Netherlands, 2021; pp. 141–186. [CrossRef]
7. Moltchanov, D.; Sopin, E.; Begishev, V.; Samuylov, A.; Koucheryavy, Y.; Samouylov, K. A Tutorial on Mathematical Modeling of
5G/6G Millimeter Wave and Terahertz Cellular Systems. IEEE Commun. Surv. Tutorials 2022, 24, 1072–1116. [CrossRef]
Future Internet 2023, 15, 290 14 of 15

8. Kondratyeva, A.; Ivanova, D.; Begishev, V.; Markova, E.; Mokrov, E.; Gaidamaka, Y.; Samouylov, K. Characterization of Dynamic
Blockage Probability in Industrial Millimeter Wave 5G Deployments. Future Internet 2022, 14, 193. [CrossRef]
9. Mokrov, E.; Samouylov, K. Performance Assessment and Comparison of Deployment Options for 5G Millimeter Wave Systems.
Future Internet 2023, 15, 60. [CrossRef]
10. ITU-T. SERIES Y: Global Information Infrastructure, Internet Protocol Aspects, Next-Generation Networks, Internet of Things and
Smart Cities; Technical Recommendation (TR) Y.3651; ITU Telecommunication Standardization Sector (ITU-T): Geneva, Switzer-
land, 2018.
11. 3GPP. 5G System (5GS); Study on Traffic Characteristics and Performance Requirements for AI/ML Model Transfer; Technical Report (TR)
22.874; Release 18, V18.2.0; 3rd Generation Partnership Project (3GPP): Valbonne, France, 2017.
12. ITU-T. SERIES Y: Global Information Infrastructure, Internet Protocol Aspects, Next-Generation Networks, Internet of Things and
Smart Cities; Technical Recommendation (TR) Y.3602; ITU Telecommunication Standardization Sector (ITU-T): Geneva, Switzer-
land, 2022.
13. Cisco. Spend Less Time Managing Your Network. 2022. Available online: https://www.cisco.com/site/us/en/products/
networking/dna-center-platform/index.html (accessed on 1 June 2022).
14. Chen, A.; Law, J.; Aibin, M. A Survey on Traffic Prediction Techniques Using Artificial Intelligence for Communication Networks.
Telecom 2021, 2, 518–535. [CrossRef]
15. Efron, B.; Hastie, T. Computer Age Statistical Inference: Algorithms, Evidence, and Data Science; Cambridge University Press:
Cambridge, UK, 2016; pp. 1–475. [CrossRef]
16. Jiang, W. Cellular Traffic Prediction with Machine Learning: A Survey. Expert Syst. Appl. 2022, 201, 117163. [CrossRef]
17. Gorshenin, A.; Kuzmin, V. Statistical Feature Construction for Forecasting Accuracy Increase and Its Applications in Neural
Network Based Analysis. Mathematics 2022, 10, 589. [CrossRef]
18. Downey, A.; Loukides, M.; Blanchette, M.; Demarest, R. Think Stats: Exploratory Data Analysis; O’Reilly Media: Sebastopol, CA,
USA, 2014; pp. 1–223.
19. Gorshenin, A.; Shcherbinina, A. Efficiency of the Method for Detecting Normal Mixture Signals with Pre-Estimated Gaussian
Mixture Noise. Pattern Recognit. Image Anal. 2020, 30, 470–479. [CrossRef]
20. Gorshenin, A.; Kazakov, I.; Korolev, V. On the Convergence of Median Versions of the Expectation-Maximization Algorithm for
the Separation of Finite Normal Mixtures. J. Math. Sci. 2022, 267, 92–98. [CrossRef]
21. Xu, F.; Lin, Y.; Huang, J.; Wu, D.; Shi, H.; Song, J.; Li, Y. Big Data Driven Mobile Traffic Understanding and Forecasting: A Time
Series Approach. IEEE Trans. Serv. Comput. 2016, 9, 796–805. [CrossRef]
22. Stepanov, N.; Alekseeva, D.; Ometov, A.; Lohan, E. Applying Machine Learning to LTE Traffic Prediction: Comparison of Bagging,
Random Forest, and SVM. In Proceedings of the 12th International Congress on Ultra Modern Telecommunications and Control
Systems and Workshops, ICUMT 2020, Brno, Czech Republic, 5–7 October 2020; pp. 119–123. [CrossRef]
23. Ma, T.; Antoniou, C.; Toledo, T. Hybrid Machine Learning Algorithm and Statistical Time Series Model for Network-Wide Traffic
Forecast. Transp. Res. Part Emerg. Technol. 2020, 111, 352–372. [CrossRef]
24. Lens Shiang, E.; Chien, W.C.; Lai, C.F.; Chao, H.C. Gated Recurrent Unit Network-based Cellular Traffic Prediction. In Proceedings
of the 34th International Conference on Information Networking, ICOIN 2020, Barcelona, Spain, 7–10 January 2020; pp. 471–476.
[CrossRef]
25. Zhaowei, Q.; Haitao, L.; Zhihui, L.; Tao, Z. Short-Term Traffic Flow Forecasting Method with M-B-LSTM Hybrid Network. IEEE
Trans. Intell. Transp. Syst. 2022, 23, 225–235. [CrossRef]
26. Shan, M.; Yan, Q.; Huang, S.; Wang, Y. Prediction and Analysis of Telemetry Data Based on LSTM Network. In Proceedings of the
2nd International Conference on Computer Network, Electronic and Automation, ICCNEA 2019, Xi’an; China, 27–29 September
2019; pp. 155–159. [CrossRef]
27. Syam, R.F.; Girsang, A.S. Bandwidth Provisioning for 4G Mobile Network Using Hybrid ARIMA-LSTM Based Traffic Forecasting.
Int. J. Eng. Trends Technol. 2021, 69, 235–241. [CrossRef]
28. Azari, A.; Salehi, F.; Papapetrou, P.; Cavdar, C. Energy and Resource Efficiency by User Traffic Prediction and Classification in
Cellular Networks. IEEE Trans. Green Commun. Netw. 2022, 6, 1082–1095. [CrossRef]
29. Tran, Q.T.; Hao, L.; Trinh, Q.K. Cellular Network Traffic Prediction Using Exponential Smoothing Methods. J. Inf. Commun.
Technol. 2019, 18, 1–18. [CrossRef]
30. Peng, Y.; Lei, M.; Li, J.B.; Peng, X.Y. A Novel Hybridization of Echo State Networks and Multiplicative Seasonal ARIMA Model
for Mobile Communication Traffic Series Forecasting. Neural Comput. Appl. 2014, 24, 883–890. [CrossRef]
31. Kurri, V.; Raja, V.; Prakasam, P. Cellular Traffic Prediction on Blockchain-Based Mobile Networks Using LSTM Model in 4G LTE
Network. Peer-to-Peer Netw. Appl. 2021, 14, 1088–1105. [CrossRef]
32. Oduro-Gyimah, F.K.; Boateng, K.O. Using Autoregressive Integrated Moving Average Models in the Analysis and Forecasting of
Mobile Network Traffic Data. J. Eng. Res. 2019, 7, 1–9. [CrossRef]
33. Céspedes, J.E.S.; Rodríguez, Y.G.; Sarmiento, D.A.L. Development of An Univariate Method for Predicting Traffic Behaviour in
Wireless Networks through Statistical Models. Int. J. Eng. Technol. 2015, 7, 27–36.
34. Bastos, J.A. Forecasting the Capacity of Mobile Networks. Telecommun. Syst. 2019, 72, 231–242. [CrossRef]
35. Ak, E.; Canberk, B. Forecasting Quality of Service for Next-Generation Data-Driven WiFi6 Campus Networks. IEEE Trans. Netw.
Serv. Manag. 2021, 18, 4744–4755. [CrossRef]
Future Internet 2023, 15, 290 15 of 15

36. Sone, S.P.; Lehtomäki, J.J.; Khan, Z. Wireless Traffic Usage Forecasting Using Real Enterprise Network Data: Analysis and
Methods. IEEE Open J. Commun. Soc. 2020, 1, 777–797. [CrossRef]
37. Shayea, I.; Alhammadi, A.; El-Saleh, A.A.; Hassan, W.H.; Mohamad, H.; Ergen, M. Time Series Forecasting Model of Future
Spectrum Demands for Mobile Broadband Networks in Malaysia, Turkey, and Oman. Alex. Eng. J. 2022, 61, 8051–8067. [CrossRef]
38. Gijón, C.; Toril, M.; Luna-Ramírez, S.; Marí-Altozano, M.L.; Ruiz-Avilés, J.M. Long-Term Data Traffic Forecasting for Network
Dimensioning in LTE with Short Time Series. Electronics 2021, 10, 1151. [CrossRef]
39. Li, Y.; Wang, Y. Mobile Virtual Reality Rail Traffic Congestion Prediction Algorithm Based on Convolutional Neural Network.
Mob. Inf. Syst. 2022, 2022, 2174208. [CrossRef]
40. Biernacki, A. Traffic Prediction Methods for Quality Improvement of Adaptive Video. Multimed. Syst. 2018, 24, 531–547.
[CrossRef]
41. Yu, Q.; Jibin, L.; Jiang, L. An Improved ARIMA-Based Traffic Anomaly Detection Algorithm for Wireless Sensor Networks. Int. J.
Distrib. Sens. Netw. 2016, 2016, 9653230. [CrossRef]
42. Feng, H.; Shu, Y.; Ma, M. WLAN Traffic Prediction Using Support Vector Machine. IEICE Trans. Commun. 2009, E92-B, 2915–2921.
[CrossRef]
43. Yadav, R.K.; Balakrishnan, M. Comparative Evaluation of ARIMA and ANFIS for Modeling of Wireless Network Traffic Time
Series. Eurasip J. Wirel. Commun. Netw. 2014, 2014, 15. [CrossRef]
44. Arifin, A.S.; Habibie, M.I. The Prediction of Mobile Data Traffic based on the ARIMA Model and Disruptive Formula in Industry
4.0: A Case Study in Jakarta, Indonesia. Telkomnika (Telecommun. Comput. Electron. Control) 2020, 18, 907–918. [CrossRef]
45. Box, G.; Jenkins, G.; Reinsel, G.; Ljung, G. Time Series Analysis: Forecasting and Control; Wiley: Hoboken, NJ, USA, 2015; pp. 1–712.
46. Cryer, J.; Chan, K. Time Series Analysis: With Applications in R; Springer: Berlin/Heidelberg, Germany, 2008; pp. 1–491.
47. Faverjon, C.; Berezowski, J. Choosing the Best Algorithm for Event Detection Based on the Intend Application: A Conceptual
Framework for Syndromic Surveillance. J. Biomed. Inform. 2018, 85, 126–135. [CrossRef] [PubMed]
48. Hyndman, R.; Athanasopoulos, G. Forecasting: Principles and Practice; OTexts: Melbourne, Australia, 2021; pp. 1–442.
49. Miao, D.; Qin, X.; Wang, W. The Periodic Data Traffic Modeling based on Multiplicative Seasonal ARIMA Model. In Proceedings
of the 6th International Conference on Wireless Communications and Signal Processing, WCSP 2014, Hefei, China, 23–25 October
2014. [CrossRef]
50. Cleveland, R.B.; Cleveland, W.S.; McRae, J.E.; Terpenning, I. STL: A Seasonal-Trend Decomposition Procedure Based on Loess. J.
Off. Stat. 1990, 6, 3–73.
51. Kwiatkowski, D.; Phillips, P.; Schmidt, P.; Shin, Y. Testing the Null Hypothesis of Stationarity Against the Alternative of a Unit
Root. How Sure are We that Economic Time Series Have a Unit Root? J. Econom. 1992, 54, 159–178. [CrossRef]
52. Efrosinin, D.; Kochetkova, I.; Stepanova, N.; Yarovslavtsev, A.; Samouylov, K.; Valentini, R. The Fourier Series Model for
Predicting Sapflow Density Flux based on TreeTalker Monitoring System. Lect. Notes Comput. Sci. 2020, 12526, 198–209.
[CrossRef]
53. Efrosinin, D.; Kochetkova, I.; Stepanova, N.; Yarovslavtsev, A.; Samouylov, K.; Valentini, R. Trees Classification based on Fourier
Coefficients of the Sapflow Density Flux. Ann. Math. Informaticae 2021, 53, 109–123. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.