Kalman Filter Models for Network Forecasting
Charles D. Pack, B. A. Whitaker · Bell System Technical Journal · 1982
The Bell System has recently completed studies that are expected to result in substantially improved forecasts for use in network planning. These improved forecasts are achieved through the use of new forecasting algorithms that employ Kalman filter models. To motivate the selection of Kalman filter forecasting procedures, we describe the Bell System's special data characteristics and processing requirements in the network planning process. We also discuss the Kalman filter models, their statistical properties, the model identification process, and certain implementation considerations.