TELK OMNIKA T elecommunication, Computing, Electr onics and Contr ol V ol. 18, No. 2, April 2020, pp. 907 918 ISSN: 1693-6930, accredited First Grade by K emenristekdikti, No: 21/E/KPT/2018 DOI: 10.12928/TELK OMNIKA.v18i2.12989 r 907 The pr ediction of mobile data traffic based on the ARIMA model and disrupti v e f ormula in industry 4.0: A case study in J akarta, Indonesia Ajib Sety o Arifin, Muhammad Idham Habibie Department of Electrical Engineering, F aculty of Engineering, Uni v ersitas Indonesia, Indonesia Article Inf o Article history: Recei v ed Apr 23, 2019 Re vised Jan 5, 2020 Accepted Feb 21, 2020 K eyw ords: Capacity planning Disrupti v e formula Industry 4.0 IoT Prediction methods ABSTRA CT Disrupti v e technologi es, which are caused by the cellular e v olution including the Inte rnet of Things (IoT), ha v e significantly contrib uted data traf fic to the mobile telecommunication netw ork in the era of Industry 4.0. These technologies cause erroneous predictions prompting mobile operators to upgrade their netw ork, which leads to re v enue loss. Besides, the inaccurac y of netw ork prediction also creates a bottlene ck problem that af fects the performance of the telecommunication netw ork, especially on the mobile backhaul. W e propose a ne w technique to predict more accurate data traf fic. This research used a uni v ariate Autore gressi v e Inte grated Mo ving A v erage (ARIMA) model combined with a ne w disrupti v e formula. Another model, called a disrupti v e formula, uses a judgmental approach based on four v ariables: Political, Economic, Social, T echnological (PEST), cost, time to mark et, and mark et share. The disrupti v e formula amplifies the ARIMA calculation as a ne w combination formula from the judgmental and statistical approach. The results sho w that the disrupti v e formula combined with the ARIMA model has a lo w error in mobile data forecasting compared to the con v entional ARIMA. The con v entional ARIMA sho ws the a v erage mobile data traf fic to be 49.19 Mb/s and 156.93 Mb/s for the 3G and 4G, respecti v ely; whereas the ARIMA with disrupti v e formula sho ws more optimized traf fic, reaching 56.72 Mb/s and 199.73 Mb/s. The higher v alues in the ARIMA with disrupti v e formula are closest to the prediction of the mobile data forecast. This result suggests that the combination of statistical and computational approach pro vide more accurate prediction method for the mobile backhaul netw orks. This is an open access article under the CC BY -SA license . Corresponding A uthor: Ajib Setyo Arifin, Department of Electrical Engineering, F aculty of Engineering, Uni v ersitas Indonesia, Depok 16424, Indonesia. Email: ajib@eng.ui.ac.id 1. INTR ODUCTION The total mobile data traf fic generated by telecommunication technologies has been significantly con- trib uting to the core netw ork in recent years. This has led to a congestion problem, especially in mobile backhaul technologies, which play a significant role in bringing traf fic to the core netw ork. If the mobile back- haul is congested, the operator performance may return a pack et drop or higher latenc y , where af fecting the end user indirectly . Besides, as the upcoming Industry 4.0 has already been introduced in se v eral countries, mobile J ournal homepage: http://journal.uad.ac.id/inde x.php/TELK OMNIKA Evaluation Warning : The document was created with Spire.PDF for Python.
908 r ISSN: 1693-6930 data traf fic no w might be dif ficult to predict. This issue can lead to erroneous predictions that might e x ert a ne g ati v e impact, such as high Capital Expenditure (CAPEX) in v estment caused by incorr ect capacity planning. Disrupti v e technologies created by inno v ati v e b usiness models ha v e changed e xisting b usiness to e v al- uate their b usiness model by adapt ing ne w technologies to monetize ne w re v enue [1][2]. According to study , the probability of creating a successful b usiness with an inno v ate approach is 10 times higher than using a non- inno v ati v e approach [3]. This kno wledge has moti v ated incumbent b usinesses to change their b usiness models according to Industry 4.0, which introduces more automation and data-e xchange systems. Therefore, in recent years, ne w or e xisting industries of fer ne w small b usiness model, simplified products and services, and ef ficient solution. This has contrib uted significantly to mobile data traf fic as well. The lar ge capacity pioneered by The Third Generation (3G) and F ourth Generation (4G) ha v e con- trib uted significantly to mobile backhaul capacity recently . Another upcoming technology is The Fifth Gen- eration (5G), which is the latest generation of telecommunication netw orking and may of ficially be launched in 2020. The 5G, which is named as Ne w Radio (NR) in radio access, has a do wnlink and uplink rate of 20 Gb/s and 10 Gb/s respecti v ely [4], according to the International T elecommunication Union (ITU) [5]. The 5G systems also ha v e w orking frequencies in the millimetre w a v e range; such minimal distances are causing infrastructure to become denser . The last and most critical f actors that might contrib ute to the netw ork in v olv e population statistics. The Indonesian population has increased by 5.6 % o v er the last v e years and is e xpected to continue to rise [6]. Accordingly , the number of people assumed to be the acti v e users of application/web services will also rise, in turn increasing traf fic in the netw ork. This can be correlated with the number of smart phones and International Mobile Subscriber Identities (IMSIs) which ha v e already pro vides much traf fic to the traf fic congestion [6]. Based on these f actors , a huge capacity is ur gently needed on the backhaul to support this kind of technology . By summing the t raf fic peak rate of these inte grated technologies, we e xpected a 2 1 Gb/s pipeline to be supported in the backhaul. Current mobile backhaul technologies, which use an outdated micro w a v e system, support around 150-300 Mb/s in each link; this should be changed to support a higher capacity by , for instance, using fiber optics or millimeter -w a v e links of up to 1 Gb/s. This paper aims to predict the forecast traf fic based on current technologies using second generation (2G), third generation (3G), fourth generation (4G), and to implement some IoT de vices to predict the short forecast for the ne xt fe w years of mix ed traf fic (2G, 3G, 4G, L TE, IoT , and 5G). Some researchers ha v e sho wn that ARIMA model has been widely used in man y t raf fic forecasting cases, for instance, determining T raf fic Channels (TCH) in GSM obtained in NMS in China [7] and determine T ime Series v alue for W i-Fi data [8]. Other references sho w that ARIMA able to be combined with dif ferent techniques to impro v e the accurac y of the predicted traf fic in the session: F or e xample, the prediction traf fic of the 802.11 W ireless Local Access Netw ork (WLAN) using Seasonal Autore gressi v e Inte grated Mo ving A v erage (SARIMA) [9]. In another case, the ARIMA model w as combined with the T ra v el Distance Algorithm (TDE) to predict road v ehicle traf fic to obtain more accurate real-time traf fic data [10]. Besides, It is also well-kno wn that ARIMA is also used in for currenc y prediction [11]. In this paper , we attempt to predict the forecast using a uni v ariate ARIMA model base d on autore- gression, mo ving a v erage, and dif ferentiated princ iples combined with disrupti v e formulas to mak e the forecast more accurate. Combining a statistical method with a judgmental technique (using disrupti v e equations) might impro v e the accurac y of the predicted traf fic, b ut sometimes this combination might leads to the erroneous trend [12, 13]. Ho we v er , research has sho wn that using a con v entional ARIMA model yields a a lar ge error rate compared with the ARIMA combination model with another technique [14, 15]. Besides, the ARIMA model is cate gorized as a short forecast, where it predicts a fe w time-series after data set. The disrupti v e formula analysis w as based on the time to mark et, v alue proposition, cost, and PEST analysis [3]. The main moti v ation of this research w as to analyze and predict the disrupti v eness of inte grated traf fic by considering a lo w cost and secure bandwidth solution. Ideally , preparing an une xpected traf fic in the radio access netw ork mobile operators should i n v est a lar ger pipeline. But these operators are mostly sa ving in v est- ment cost in the radio access netw ork. Therefore, this research attempted to predict the traf fic in the mobile backhaul precisely with the cost and bandwidth consideration. This paper is or g anized as follo ws: Section 2 gi v es a literature re vie w of forecasting methods that sup- port a basic understanding of this research. Section 3 e xplains the methodological points of traf fic forecasting that were used in this research. Section 4 and 5 present the results and discussion of this research, respecti v ely , including the analysis and some findings. The last section pro vides a conclusion of this w ork. TELK OMNIKA T elecommun Comput El Control, V ol. 18, No. 2, April 2020 : 907 918 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA T elecommun Comput El Control r 909 2. FORECASTING METHODS Se v eral techniques ha v e been proposed to predict technological disrupti v eness, especially in I nd us try 4.0. The first technique is called F orw ard-Citation Node P air Algorithm (FCN A), which w as introduced by Changw oo Choi and Y ongtae P ark [16], and uses a patent-citation matrix consisting of a set of nodes con- nected by arcs. This technique aims to identify the main de v elopment path of the comple x patent-citation by understanding both present and past technologies [17], where the leading technology possesses the main patents that are link ed to the selected arcs. The other technique, to impro v e FCN A, is K-core analysis, which concentrates on the sub-groups’ nodes rather than on the main patents [17]. This technique aims to remo v e the central patents (which are a ssumed to be the most disrupti v e technologies) by distrib uting them into dif ferent subgroups that can help identify the essential data [18]. The last technique to identify disrupti v eness in Industry 4.0 is called topic modelling and is similar to a search engine. T opic modelling is mostly used in the cluster that has been defined in the K-core analysis. The highest number of repetiti v e w ords in each cluster defined in K-core analysis leads to the most important aspects of the te chno l ogy . Afterw ard, to v alidate the results raised by this technique, it is recommended that tw o e xperts re vie w them [17]. All these techniques are mostly aimed at identifying the most important disrupti v e technologies in the mark et through clustering. Such techniques help mark et leaders identify which technologies are more disrupti v e, b ut the y do not determine ho w much traf- fic each will contrib ute. Based on the analysis of these techniques, the major contrib ution technologies are IoT , artificial intelligence, financial technology (including blockchain), virtual reality , and autonomous v ehicles. 2.1. T ypes of f or ecasts Judgmental methods are based on intui tion, personal i nterest, and user e xperiences [19]. One e xample of a judgmental method is the Delphi method, which emplo ys a panel of e xperts to analyze research results to ensure v alidity . A judgmental method will also be used in this prediction to analyze the disrupti v e formula using a risk-f actor technique. Uni v ariate methods depend on past and present v alues that ha v e been forecasted in a single series [19]. Uni v ariate met h ods are used in this research to analyze the predicted traf fic forecast. The ARIMA model consists of both uni v ariate and multi v ariate models. Multi v ariate models use more than one independent v ariable (time series) simultaneously to predict the forecast. These v ariables might comprise interrelationships using a dif ferent time v ariable. 2.2. The ARIMA model ARIMA stands for Autore gressi v e Inte grated Mo ving A v erage [19]. The ARIMA model is based on the Box and Jenkins method of using three dif ferent concepts: Autore gression (AR), Mo ving A v erage (MA), and inte gration, together classified as an ARIMA( p; d; q ); p defines the AR; d defines the dif ferential; and q defines the MA. AR is a technique for analyzing the past and present v alues of a data set. AR is denoted as p , where it sho ws the weighted linear of sum p v alues based on ARIMA ( p; d; q ) terminology . The p v alue indicates the number of order . The formula to denote this AR is sho wn in (1): y t = 1 y t 1 + 2 y t 2 + : : : + p y t p + (1) where p is used to determine the number of orders of past v alues; t is the time seri es; is the slope coef fic ient of the weighted past v alues; and y is the time-series function of the AR IMA model. The error term is normally distrib uted with mean zero and v ariance 2 . The MA process is denoted by order q in the ARIMA ( p; d; q ) classification, which sho ws an error v alue in (1). The error term is normally distrib uted with mean zero and v ariance 2 . MA also uses the number of orders in the past v alues, as denoted in (2): y t = t + 1 t 1 + : : : + q t q (2) where t is the time series; is the slope coef ficient of the weighted past v alue; is the number of orders needed to identify the past v alues; and y is the time-series function of the ARIMA model. T o identify ho w man y orders are in the calculation of AR, the parameter of q is used. MA has been used for stock trading. MA aims to eliminate the noises or peaks from random noise fluctuations in the graph, where it leads to the erroneous prediction. T o calculate the a v erage v alue in the chart, MA tak es a certain period, such as se v en v alues, to calculate the a v erage as sho wn in Figure 1. The pr ediction of mobile data... (Ajib Setyo Arifin) Evaluation Warning : The document was created with Spire.PDF for Python.
910 r ISSN: 1693-6930 Figure 1. The MA concept e xample Figure 1 describes an MA concept where a v erage A, B, and C are calculated. The results of A, B, and C are 1 ; 2 : 6 ; and 2 : 6 , respecti v ely . The a v erage A shares the same a v erage v alue wit h each v alue in the sequence, which is one. Due to inconsistent sequences in series six, the a v erages B and C ha v e di f ferent v al- ues compared to A. Ho we v er , instead of using an original sequence v alue that has a significant peak, among others, the MA sho ws a smooth graph. Otherwise, Inte grated or dif ferentiated v ersions are denoted as d in ARIMA ( p; d; q ), which is defined as the parameter that checks whether the graph is stationary [20]. As a best practice case, the time-series graph is mostly non-stationary . Therefore, MA and AR are not suf ficient to deter - mine the prediction. Being non-stationary might cause problems that can lead to error prediction. Therefore, a dif ferential or inte grated model is one of the best techniques a v ailable to mak e graphs stationary [19]. 3. TRAFFIC FORECASTING FORMULA TION The ARIMA model is a statistical model that predicts the forecas t based on past and present v alues. The statistical model might be inaccurate if ne w technologies or trends are af fecting the graph, making them more disrupti v e. Since a judgmental approach can predict the future more accurately [13], this paper proposes a ne w closest prediction approach to identify traf fic usi n g a combination of judgmental and statistical approaches. T o use both approaches, three important procedures must be completed: Analysis of the data set for 3G and L TE traf fic using the ARIMA model; generation of a disrupti v e formula based on en vironmental technologies in the particular country; v alidate the disrupti v e formula combined with the ARIMA model. 3.1. Determining the data set The data set w as obtained from one of the b usiest traf fic site in Indonesia where 2G, 3G, and L TE ha v e been installed. W e took an e xample from the capital c ity of Indonesia, Jakarta, where L TE w as installed in the first month of 2017. This site is strate gically located in the capital city , where internet penetration is relati v ely greater than other locations. Besides, this site also represents the possibility t o implement the IoT and the 5G netw ork. L TE and 3G mobile data t raf fic data were collected for one year . Because of mobile operators ha v e not in v ested in 2G technology an ymore, this w as assumed to be stagnant and e xcluded in the ARIMA and disrupti v e calculations. Based on the Cisco V isual Netw orking Inde x, the penetration of 2G de vices has been decreasing gradually o v er the last v e years [21]. Therefore, we assumed that 2G mobi le traf fic consumes a maximum of 2 Mb/s at each site based on maximum capacity , where traf fic remains stable. IoT sensors were also e xcluded in the calculation because Indonesia just will launch the Narro w-band IoT License in Indonesia in the latest 2018, where IoT just of ficially starting in the ne xt year . . Ho we v er , it is predicted that 400,000 IoT sensors might be installed by 2022 [22]. This trend is similar to that for 5G systems, which are e xpected to e xpand by 2020, especially in Indonesia [23]. Therefore, this research made se v eral assumptions about these technologies (2G, 5G, and IoT), whereas 3G and L TE were more clearly demonstrated. 3.2. A new disrupti v e f ormula The peak data rate of L TE and 3G, which are 300 Mb/s do wnlink and 63 Mb/s do wnlink (with 3rd carriers) respecti v ely , ha v e not sho wn a real f act that full throughput is used. Ev en 4G and 3G ha v e 300 Mb/s and 63 Mb/s in the do wnlink, only approximately 20% of the full throughput w ould be used [24]. As a result, traf fic is unpredictable and might halt one day , depending on human beha viour and compan y profiles. This paper proposes a ne w modification to the ARIMA formula. A ne w disrupti v e formula w as defined as a judgemental method that might in v olv e e xperts to identify the resul ts. The combination of the judgemental and analytical methods can impro v e the accurac y of the prediction [13]. W e proposed a ne w formulation, to predict disrupti v eness in the future: TELK OMNIKA T elecommun Comput El Control, V ol. 18, No. 2, April 2020 : 907 918 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA T elecommun Comput El Control r 911 y t = ( 1 y t 1 + 2 y t 2 + : : : + p y t p + t + 1 t 1 + : : : + q t q ) 2 D : (3) where D depends on four v ariables: time to mark et (TTM), Cost, Politics, Economics, Social, and T echnological (PEST), and Mark et Share. The v alue of disrupti v eness ranges by 0 D 1 . The v alue of tw o and the disrupti v eness range were inspired by the Global mobile data traf fic forecast in [28], where it defines data traf fic which ne v er reached more than tw o times compared to the pre vious year e v ery year; this moti v ated the creation of Equation (3) to define the D formula, which applies to both mobile backhaul and mobile backbone traf fic. In this paper , we proposed an ARIMA model as a le g ac y formula for modification. As a consequence, the results obtained from the ARIMA model were amplified using the disrupti v eness formula based on four v ariables. As a result, the result of the time-series function returns Equation (3), which is based on the disrupti v eness and le g ac y formula. W e can immediately see that the Global mobile data traf fic forecast until 2021 increases gradually each year [25]. The a v erage increase o v er v e years is around 47.4 % . W e e xtract the incremental traf fic each year as illustrated in T able 1. The incremental traf fic each year moti v ated us to analyze more deeply ho w to determine the disrupti v e formula. T able 1 sho ws that the mobile traf fic trend ne v er reached a roun d tw o times compared to the pre vious year , which w as later defined as the maximum disrupti v eness. Ho we v er , due to unpredicted technologies, mobile data traf fic might not increase at all. Therefore, the range of disrupti v eness v alues applies from 0 to 1, where 1 defines the maximum disrupti v eness by tw o times the formula, and 0 defines the minimum disrupti v eness, which remains the same. T able 1. Incremental traf fic based on the Global mobile data traf fic forecast [25] Y ear Incremental T raf fic ( % ) 2017 54.14 2018 54.54 2019 41.18 2020 45.83 2021 40 The disrupti v eness v alue is a judgemental method based on current beha viour represented in the TTM, cost, PEST , and mark et share v ariables. These v ariables might af fect the disrupti v eness v alue i n the formula. TTM is the period during which a product has been agreed upon and resources ha v e been committed to a project. The TTM is di vided into tw o v ariables: impact and probability . The length of the TTM gi v es it fle xibility to decrease a nd increase significantly , depending on time-related processes [26]. The simpler the products or services pro vided, the shorter the TTM will be. The impact of the implementation products/services and the probability that it will be de v eloped af fect the v alue of the TTM. T able 2 sho ws ho w scoring the TTM to define the disrupti v eness formula. Cost defines the total cost, including the v ariable and fix ed costs and e v en operational and capital costs. The main general cost discussed here is the amount needed to create a ne w project [27]. The cost is di vided into tw o dif ferent v ariables: impact and probability . The cost v ariable’ s v alue ranges from 0 to 1; its definition is pro vided in T able 2. PEST is considered in this disrupti v e technique to determine the performance and acti vities of b usinesses, especially in the long term [28]. It is clear that PEST might af f ect technology , especially when the technology i s le g alized. F or e xample, a ne w technology might not be implemented in a ne w project if the rele v ant authorities do not allo w it; this will in turn af fect the implementation of the ne w technology . T able 2 sho ws the PEST correlation in this v ariable which is di vided by tw o; PEST Impact and PEST Probability . As with TTM and cost, thi s PEST v alue ranges from 0 to 1. Mark et Share v ariable represents the percentage of mark et share of an industry mark ets total sales o v er a certain period. The mark et s h a re is v ery important to determining the le v el of competiti v eness among competitors. T able 2 sho ws that the mark et share w as di vided into tw o v ariables: mark et share impact and mark et share probability . The mark et share ranges from 0 to 1; each definition is sho wn in T able 2. The formula used to identify disrupti v eness incorporates the four v ariables as sho wn in T able 2. These four v ariables were used to identify the v alue of the disrupti v eness, as sho wn in (3). Each v ariable has its o wn weight which af fect the disrupti v eness formula in (4). The pr ediction of mobile data... (Ajib Setyo Arifin) Evaluation Warning : The document was created with Spire.PDF for Python.
912 r ISSN: 1693-6930 D = 4 Cost + 3 TTM + 2 PEST + Mark et Share 10 Impact + 4 Cost + 3 TTM + 2 PEST + Mark et Share 10 Probability (4) The range of the disrupti v eness v alue is defined in T able 2. Each v ariable has its o wn weight or priority , as sho wn in T able 2. The priority of e v ery v ariable such as Mark et Share, PEST , TTM, and Cost ha v e dif ferent v alues as sho wn in T able 2, which will af fect the formula of disrupti v eness in (4). It is assumed that the v ariable of Cost leads as a first priority to af fect disrupti v eness, whereas mark et share has the lo west priority . The first or last priority identify the weight v alue in each v ariable which leads t o the final formula of disrupti v eness, as e xpressed in (4). T able 2. The risk f actor and v ariables in detail 2*V ariable 2*Priority 2*Disrupti v eness Risk F actor Impact Probability 2*TTM 2*2 Score 1 TTM runs shorter as the impact of the implementation product/services is LARGE The probability that a SHOR T TTM for these products/services are implemented in the netw ork Score 1 TTM runs longer as the impact of the implementation product/services is SMALL The probability that a LONG TTM for these products/services are implemented in the netw ork 2*Cost 2*1 Score 1 T otal Cost compared to the impact of the V alue Proposition. HIGH means still af- fordable The probability that the products will sell to the customer . HIGH means sti ll af- fordable Score 0 T otal Cost compared to the impact of the V alue Proposition. LO W means rela- ti v ely not af fordable The LO W probability of the technology to of fer support in terms of PEST 2*PEST 2*3 Score 1 Ha ving a LARGE impact on PEST for supporting these technologies The HIGH probability of the technology to of fer support in terms of PEST Score 0 Ha ving a SMALL impact on PEST for supporting these technologies The LO W probability of the technology to of fer support in terms of PEST 2*Mark et Share 2*4 Score 1 The HIGH impact of the mark et share on subscribers The percentage of mark et share to lead others Score 0 The LO W impact of mark et share on subscribers The percentage of mark et share to lead others 3.3. F or ecasting err or management The analysis of this formula compares the global mobile data forecast prediction with the model calculation used in this paper . The disrupti v e formula w as analyzed using a percentage error comparing between con v entional ARIMA and ARIMA with disrupti v e formula, which are defined as follo ws: % E r r or = x y y 100% : (5) where x refers to the Global mobile data traf fic forecast , and y refers to the type of model used–in this case, the con v entional ARIMA and the ARIMA with disrupti v e formula. The percentage error defined the significance of an error compared to the mobile data traf fic, where the lo west error rate led to better performance in deciding the forecast v alue. 4. 3G AND 4G FORECASTING The formula of disrupti v eness combined with ARIMA has been defined. This section will e xplain tw o main analyses: 3G and 4G forecasting. Based on the mobile dat a traf fic data set in Figure 2, the 3G traf fic reached the highest payload traf fic in September 2016. Ho we v er , during the subsequent year , 3G traf fic reduced slo wly in March 2017, and remains stable afterw ards. TELK OMNIKA T elecommun Comput El Control, V ol. 18, No. 2, April 2020 : 907 918 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA T elecommun Comput El Control r 913 Figure 2. Comparison between ARIMA model and ARIMA with disrupti v e formula Based on the ARIMA calculation, defined in the black line in Figure 2, the prediction of this tr af fic sho ws a small decrease in the first month, after which it remained stable until the end of March 2019. The ARIMA calculation used an ordered ARIMA (2,1,3), where it defined order tw o for AR, one for dif ferentiating, and three for MA. The order calculation is calculated by data analysis tools, whi ch i s called R studio, to support predicti v e analysis using a Akaik e Information Criterion (AIC). Based on the combined ARIMA with a disrupti v eness formula, as sho wn in the dotted line in Figure 2, the graph sho ws a significant v alue on ARIMA with a disrupti v e formula compared to the con v entional ARIMA. The disrupti v eness v alue w as based on the cost, TTM, PEST , and mark et share v ariables, which are defined in T able 3. T able 3. The disrupti v e v ariables in the 3G and 4G Netw orks 2*V ariable 3G Netw ork 4G Netw ork Impact Probability Impact Probability TTM 0.2 0.2 0.5 0.7 Cost 0.4 0.8 0.7 0.5 PEST 0.3 0.8 0.5 0.5 Mark et Share 0.6 0.6 0.7 0.7 D 0.34 0.6 0.6 0.58 Based on the results in Figure 2, there w as a dif ference in traf fic of around 8 Mb/s o v er the year . W e conclude that from 2018 to 2019, based on T able 3, the PEST and cost v ariables, especially with respect to probability , sho wed a significant v alue, which reached 0.8 from 1.0. This might ha v e caused the 3G mobile netw ork penetration, which is still promised for technology implementation across the Indonesian islands. In f act, Base Station (BS) are still una v ailable on se v eral islands especially rural areas. Therefore, 3G might be preferred for application compared to other technologies re g arding PEST and cost probability . Ho we v er , PEST might not af fect the o v erall disrupti v eness v alues much, since it is a third priority , after cost and TTM. The TTM v ariable illustrated in T able 3 sho ws a small v alue of 0.2 from 1.0, which w as caused by the 3G netw ork trend in 2019. Since mobile netw ork t rends are mo ving to w ards L TE and 5G netw orks in 2019 to support lo w-po wer de vices, 3G might not be preferable for implementation in mobile traf fic. Figure 3 sho ws from the trends in the data set that L TE mobile data traf fic significantly increased from March 2017 until the end of No v ember 2017, b ut then remained stable until March 2018. The prediction sho ws constant data traf fic o v er a year starting f rom March 18. Based on tw o calculations, the trend prediction analysis using ARIMA and the ARIMA with disrupti v e formula as depicted in the graph in Figure 4 after March 2018, sho ws dif ferent payload traf fic around 40 Mb/s o v er a year . Figure 3. Comparison of 4G mobile traf fic with a prediction The pr ediction of mobile data... (Ajib Setyo Arifin) Evaluation Warning : The document was created with Spire.PDF for Python.
914 r ISSN: 1693-6930 Based on ARIMA as sho wn in the black line in Figure 3, the prediction traf fic sho ws a small decrease at first and then remains stable o v er a year . The ARIMA model used in this formula used ARIMA (1,1,1) based on the AIC calculat ion in R Studio. Based on ARIMA with disrupti v e formula in the dotted line from Figure 3, the ef fects of four v ariables introduced in Section IV .B sho w a significant contrast with the ARIMA formula. The dif ference in v alue between ARIMA with disrupti v e formula and the ARIMA model is approximately 40 Mb/s o v er the time series. During this year , the four v ariables mostly had the same a v erage for the Impact and Probability . In the 4G netw ork, based on T able 3, the mark et share leads in 4G using a relati v e higher v alue compared to other v ariables, which is 0.7 for both probability and impact. This is mainly caused by 4G netw ork penetration, which increased relati v ely from 2018 to 2019. The TTM Probability and cost Impact had a v alue of 0.7 from 1.0, since 4G is an af fordable technology to support higher traf fic. A higher cost for 4G might still be preferable if compared to the impact of this technology on users, where most people and some de vices are using more data than in pre vious years. This might cause the cost impact to be higher than others. The TTM Probability , sho wing 0.7 from 1.0 in T able 3, w as caused by the impact of this technology as well. The beha viour of people in 2019 is e xpected to support digitalization technology that consumes more data traf fic in the netw ork. This will lead to shorter TTM to support the mobility de vices in the netw ork. 5. RESUL T AND DISCUSSION The simulation results using ARIMA model and an ARIMA combination with disrupti v e formula ha v e been described in Figure 2 and Figure 3, respecti v ely . The increasing traf fic using a disrupti v eness formula for 3G and 4G technologies significantly escalated the base v alue of ARIMA model by 8 Mb/s and 40 Mb/s, respecti v ely . The four v ariables were deemed more promising for accurate prediction compared with using the ARIMA model. The ARIMA calculation, which w as based on past and present v alues and MA, might generate inaccurate predictions if disrupti v e technologies are not considered. T o assess this issue, this study utilised a percentage error that compared between con v entional ARIMA and ARIMA with dis rupti v e formula with the global mobile data traf fic forecast. 5.1. Err or perf ormance The Error performance subsection aims to compare the percentage error between tw o models in 3G and 4G traf fic based in the results obtained in Section 4. Based on (5), the global data traf fic v ariable used a global mobile data forecast [28], which w as identified using a data set multiplied by the incremental v alue in T able 1 in the year 2018-2019. Additionally , the model calculation in (5) applied the con v entional ARIMA and ARIMA with disrupti v e formula. The global mobile data traf fic calculated the a v erage mobile data traf fic obtained from the data set in 3G and 4G—i n this case, from 2016 to 2018. The data set w as a v eraged o v er 2016 to 2018 and multiplied using an incremental v alue based on T able 1, which w as 41.17 % from 2018-2019. By obtaining the data set traf fic and the incremental v alue from this data, the optimized prediction traf fic w as obtained, which is sho wn in T able 4. T able 4 sho ws optimized traf fic based on Mobile Data F orecast, where it increases 41.17 % from 2018 to 2019. The data traf fic in 2019 based on T able 4 will be compared to the prediction based on Con v entional ARIMA and ARIMA + Disrupti v e F ormula. It assumed that the global mobile data traf fic forecast has more accurate prediction based on se v eral types of research. T able 4. T raf fic comparison between 3G and 4G in 2018-2019 2*T echnology Y ear 2018 2019 3G 55.06 Mb/s 77.63 Mb/s 4G 207.87 Mb/s 293.09 Mb/s In 3G, the ARIMA with disrupti v e formula reaches 56.67 Mb/s, which is almost optimized to 72.17 Mb/s in T able 4. The error rate seemed to be lo wer than in the con v entional ARIMA. Besides, in 4G, the ARIMA with disrupti v e formula reaches 199.6 Mb/s, whereas con v entional ARIMA reaches 156.93 Mb/s, which is f ar higher than the global mobile data traf fic, 293.09 Mb/s in T able 4. The error rate in both v al- ues is caused by se v eral f actors, i.e.: backhaul traf fic and de v eloping country f actors. The calculation of the ARIMA model used backhaul traf fic, where it seemed to be more ne g ati v e compared to the global data forecast. TELK OMNIKA T elecommun Comput El Control, V ol. 18, No. 2, April 2020 : 907 918 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA T elecommun Comput El Control r 915 Ho we v er , the lo wer error rate in ARIMA with disrupti v e formula is more promising compared to the con v entional ARIMA, where it defined more optimized v alues from 2018 to 2019. This demonstrates that the ARIMA with disrupti v e formula had more accurate prediction compared to the con v entional ARIMA. Internet penetrat ion in de v eloping countries has increased relat i v ely slo wly compared to de v eloped countries, which might af fect mobile data traf fic. In this case, compar ed to global mobile data traf fic, the con v entional ARIMA sho ws a small increase o v er a year for 3G and 4G, as sho wn in Figures 3 and 4. Ho we v er , the ARIMA with disrupti v e formula w as more positi v e, which is closely related to the mobile data traf fic trend and the internet implementation program across Indonesia. This led to more accurate prediction. 5.2. V ariable analysis The four f actors that ha v e been defined-TTM, Cost, PEST , and Mark et Share-significantly af fected the con v entional ARIMA. As e xplained earlier , the v ariable cost had maximum priority o v er others, while mark et share w as less important. T o v alidate the v ariables, we identified the dif ferent le v els of each v ariable using the maximum incremental traf fic of Impact and Probability . The maximum order of Impact and Probability , M , can be e xpressed, M = N max ( I mpact P r obabil ity ) : (6) where N is equal to the number of weight, which is identified in T able 2. F or e xample, the v ariable cost has four maximum weights, where the v ariable mark et share has 1 maximum weight. Besides, Impact and Probability are assumed to be at maximum probability , which is equal to 1. It also assumes that the other v ariables are zero if a particular v ariable is calculated. Based on (6), T able 5 sho ws dif ferent maximum orders in each v ariable. Moreo v er , Figure 4 took an e xample in the 3G netw ork, sho wing the dif ferent incremental traf fic using the four v ariables identified in T able 5. The incremental traf fic sho ws a dif ferent prediction w as used in the ARIMA model. T able 5. Maximum incremental traf fic Disrupti v e V ariables Incremental T raf fic ( % ) TTM 9 Cost 16 PEST 4 Mark et Share 9 Figure 4. Comparison of the dif ferent mobile netw orks Based on Figure 4 and T able 5, the v ariable cos t will af fect a maximum 16 % incremental v alue compared t o the con v entional ARIMA. Cost is cross-related to the re v enues of companies. This is reasonable since if we imagine that companies ha v e a significant amount of re v enues to utilize the initial and maintenance costs for Capital Expenditure (CAPEX) and Operational Expenditure (OPEX), the y will prioritize customer demands, including data traf fic speed and latenc y , which are essential for users. By this v ariable, in the Indonesian case, the operators ha v e an opportunity to either spread the mobile base stations into dif ferent locations or to mak e re gular netw orks denser to increase their capacity . The cost might correlate with other v ariables, such as PEST , TTM, and Mark et Share. If mobile operators h a v e more re v enues, this will af fect disrupti v eness and other v ariables. F or e xample , the re v enues will consider the frequenc y allocation that has been determined by the re gulators in each particular country , where more re v enues will probably decide more frequenc y allocations. The pr ediction of mobile data... (Ajib Setyo Arifin) Evaluation Warning : The document was created with Spire.PDF for Python.
916 r ISSN: 1693-6930 TTM and PEST are the second and third pri orities in the disrupti v eness v ariables. The maximum total incremental traf fic is 16 % and 9 % , respecti v ely , as sho wn in Figure 4. The e xample real-w orld case of this v ariable is the license readiness to implement ne w frequencies and technologies i n the country . F or e xample, the main challenge for the license readiness e xample is the millimetre w a v e in 5G technology , where se v eral steps are needed to assess the frequenc y spectrum in their country . The ne w spectrum in millimetre w a v e should consider permission to open a ne w license spectrum. This license is also a dependent f actor with the cost profile of the mobile operators, where ha ving a ne w spectrum leads to higher risk of spending at more considerable cost. Besides, technology readiness is also a dependent f actor with the cost profile of mobile companies, where technology readiness might be delayed if traf fic penetration in the country is not co v ered 100 % . As an e xample, in Indonesia, the L TE mobile stations will be implemented later , since 3G stations ha v e not yet been implemented across the whole of Indonesia. Therefore, to support more ef ficienc y , 3G is still preferable to L TE, which reduces more CAPEX and OPEX, for ne w stations. Both cases might cause a delay in technology implementation for the country , where both v ariables still depend mostly on cost. As a result, cost is still the highest priority af fecting mobile data traf fic. Mark et share, the lo west priorit y , determines only 1 % of the incremental traf fic. Mark et share does not af fect mobile data traf fic v ery much if traf fic and re v enues are relati v ely increasing in v ersely , which has occurred in the telecommunication en vironment. This mak es mark et share the lo west priority in the disrupti v eness v alue. T o conclude, four v ariabl es are the main f actors of disrupti v e traf fic, with cost/re v enues being the most dominant f actors that af fect disrupti v e traf fic. 5.3. Backhaul analysis The arri v al of disrupti v e technologies will af fect the total capacities in the mobile backhaul. As sho wn in Section 5, 4G and 3G traf fic ha v e amplified traf fic in the con v entional ARIMA, around 35 Mb/s and 10 Mb/s, respecti v ely . Other technology , such as 2G systems, which is assumed to be stagnant, consumes around 2 Mb/s each year . Besides, the 5G netw ork has not been implemented yet, and while IoT systems are increasing this year , we still assumed no IoT due to ha ving the lo west bit rate and smallest total number of de vices. Illustrating all these f actors, T able 6 presents an o v ervie w of the predicted a v erage mobile data traf fic between 2017 and 2018, mainly in each site, using ARIMA with disrupti v e formula. T able 6. A v erage Mobile Data T raf fic 2017 and 2018 2*T echnologies A v erage Mobile Data T raf fic(Mb/s) 2018 2019 2G 2 2 3G 54.62 56.67 4G 198.03 199.6 T otal 254.65 258.23 By calculating the ARIMA with disrupti v e formula model, we conclude that the mobile backhaul could support the old micro w a v e technologies, where the future backhaul will need at least the a v erage of 258.2 Mb/s each site based on T able 6. This capacity is basically could be supported by the e xisting micro w a v e technologies. Ho we v er , to anticipate the une xpected traf fic in the future, this paper recommends a list of feature for the mobile backhaul, which are (from the most ef ficient): Using HOM that supports traf fic greater than 300 Mb/s, for instance, with 1024 or 2048 QAM, implementing more antennas in the mobile backhaul systems to increase capacity , such as MIMO or Massi v e MIMO, migrating old micro w a v e technologies to fiber optics. This paper founds an ef fecti v e and accurate w ay to predict the traf fic forecast based on statistical and judgemental approach. W ith the combination of ARIMA model and disrupti v e formula that this paper proposed, it has sho wn that ARIMA is more accurate if it is associated with a Judgemental approach to correct the errors. 6. CONCLUSION The major contrib ution of the study is the de v elopment of a ne w formula in the ARIMA model to predict forecast traf fic based on four v ariables: TTM, cost, PEST , and mark et share. Our research confirms that disrupti v e technology af fect the mobile data traf fic if: the telecommunication companies are profitable; the time to mark et to implement ne w project is acceptable, the en vironment of PEST supports ne w technologies TELK OMNIKA T elecommun Comput El Control, V ol. 18, No. 2, April 2020 : 907 918 Evaluation Warning : The document was created with Spire.PDF for Python.