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