Page 102 - ITU Journal - ICT Discoveries - Volume 1, No. 2, December 2018 - Second special issue on Data for Good
P. 102

ITU JOURNAL: ICT Discoveries, Vol. 1(2), December 2018





                                            * 4XDUWHUO\ 'RZQOLQN 'DWD 7UDIILF

                    30000000                                                                      Actual
                   Data Traffic (kbps)  10000000 0  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15
                                                                                                  CANFIS
                    20000000




                                                     Time (Quarters)


                              Fig. 7 – Actual and predicted 3G quarterly traffic using the CANFIS (5-2-5) model

          4.   CONCLUSION                                      [4]   Purnawansyah  and  Haviluddin,  "Comparing
                                                                     performance  of  backpropagation  and  RBF
          The  CANFIS  model  has  been  used  to  predict  five     neural  network  models  for  predicting  daily
          time  spans  of  hourly,  daily,  weekly,  monthly  and    network    traffic",   The  4     Makassar
                                                                                                  th
          quarterly 3G downlink data simultaneously. In this         International  Conference   on    Electrical
          approach  five  different  CANFIS  models  were            Engineering   and   Informatics   (MICEEI),
          developed  and  the  CANFIS  (5-2-5)  model  was           Makassar City, Indonesia, pp. 166-169, 2014.
          selected as the best. The model was evaluated by
          comparing  the  forecast with  actual  data  obtained   [5]   M.  D.  Junior,  J.  D.  Gadze,  and  D.  K.  Anipa,
          from 3G mobile operator and the results showed a           "Short-term  traffic  volume  prediction  in
          good  performance  with  minimum  values  of  MSE,         UMTS  networks  using  the  Kalman  filter
          NRMSE and percent error of 0.000486, 0.01120 and           algorithm",  International Journal  of  Mobile
          12.33%.                                                    Network Communications & Telematics, vol. 3,
                                                                     no. 6, pp. 31-40, 2013.
          In  the  future,  a  genetic  algorithm  optimization
          technique will be explored to improve on the delay   [6]   X.  Dong,  W.  Fan,  and  J.  Gu,  “Predicting  LTE
          in training of the CANFIS model when membership            throughput  using  traffic  time  series”,  ZTE
          function  per  input  and  multiple  input  data  are      Communications,  vol. 13, no. 4,  pp. 61-64,
          increased.                                                 December 2015.


          REFERENCE                                            [7]   G. Pandey, K. M. Siddiqui and A. K. Choudhary,
                                                                     “Telecom  voice  traffic  prediction  for  GSM
          [1]   Y.  Yu,  M.  Song,  Y.  Fu  and  J.  Song,  “Traffic   using   feedforward   neural   network”,
               prediction in  3G mobile networks  based on           International Journal of Engineering Science
               multifractal  exploration”,  Tsinghua Science         and Technology,  vol.  5  no.  3,  pp.  505-511,
               and Technology,  vol.  18,  no.  4,  pp.  398-405,    March 2013.
               August 2013.
                                                               [8]   P.  Svoboda,  M.  Buerger,  and  M.  Rupp,
          [2]   B.  Yang,  W.  Guo,  Y.  Jin  and  S.  Wang,         “Forecasting of traffic load in a live 3G packet
               “Smartphone  data  usage:  downlink  and              switched  core  network”,  5  International
                                                                                                 th
               uplink  asymmetry”,  Electronics Letters,             Symposium on  Communication Systems,
               vol. 52, no. 3, pp. 243-245, February 2014.           Networks and Digital Signal Processing,
                                                                     CNSDSP 2008, Graz, Austria, 25 July, 2008.
          [3]   X.  Dong,  W.  Fan  and  J.  Gu,  “Predicting  LTE
               throughput using traffic timeseries”,  ZTE      [9]   I. A. Lawal, S. A. Abdulkarim, M. K. Hassan, and
               Communications,  vol. 13, no. 4,  pp.  61-64          J.  M.  Sadiq,  “Improving  HSDPA  traffic
               December 2015.                                        forecasting  using  ensemble  of  neural
                                                                     networks”, 15  IEEE International Conference
                                                                                  th
                                                                     on Machine Learning and Applications
                                                                     (ICMLA),  Anaheim,  USA,  18-20  December,
                                                                     2016.





         80                                  © International Telecommunication Union, 2018
   97   98   99   100   101   102   103   104   105   106   107