Prediksi Trafik Data pada Jaringan Telekomunikasi Menggunakan Metode Jaringan Saraf Tiruan Propagasi Balik

Abstract

The development of communication technology has had a significant impact on the increase of communication service user. As a result, there is a high traffic on UMTS networks. In order to maintain the network’s Quality of Service (QoS), PT Indosat guarantees that its network utilization will not exceed 80%. This research tries to predict the future traffic, so that the company can analize the requirement of additional network capacity. The utilited data is the time series data of packet switching (PS) traffic on node B IPB, from January 2011 until April 2012. The used traffic data are original data, differentiated data, and normalized data. The normalized data are obtained by three functions, namely: logarithmic function, min-max normalization, and z-index normalization. Artifiacial neural network (ANN) back propagations used as the prediction method. The prediction error is calculated using mean absolute persentage error (MAPE). The result of this research is a three-month prediction of traffic data starting from April 2012. The prediction result with the smallest error of 5.14% is obtained from the data that are already transformed with logarithmic function. This small value of MAPE shows that the prediction result is quite accurate.