A short-term forecasting algorithm for trunk demand servicing

Trunk servicing is the continual process of collecting trunk group traffic measurements, monitoring network service, and augmenting the network when necessary. This study addresses the possibility of using a short-term forecast to determine the adequacy of trunk quantities planned for the imminent busy season. When seasonal patterns of demand exist, it may be possible to use observed, pre-busy-season traffic levels to predict accurately that busy-season demand will exceed the planned trunk group capacity and to determine appropriate corrective action. Toward this end, we develop a seasonal load forecasting algorithm based on Kalman filter estimation techniques and analyze the effectiveness of this approach using Bell operating company data. For trunk groups exhibiting seasonal demand, the short-term (11/2 months ahead), seasonal forecast error is 50 percent less than that of the sequential projection algorithm (SPA), which linearly trends the yearly busy-season loads. Much of this improvement is attributed to the ability of the seasonal algorithm to utilize recent observations; the one-year ahead seasonal forecast error is only 20 percent less than that of SPA. We conclude that the greater generality and simplicity of SPA makes that algorithm the appropriate choice for the annual busy-season trunk forecast used in medium-range network planning. However, the seasonal algorithm demonstrated the ability to use recent data to respond quickly and accurately to various situations that result in inaccurate SPA forecasts. For this reason, the short-term forecasting algorithm developed herein is a potentially valuable tool for network administration.