Packet loss rate prediction using a universal indicator of traffic

In a multimedia environment, prediction of the quality of service plays an important role in formulating traffic control functions. We discuss a new approach in predicting the packet (or cell) loss rate as the quality of service of interest. While the approach does not rely on an assumption of a statistical model for the traffic patterns, it closely approximates the actual cell loss rate in a multimedia environment. To do this, first, we identify a set of traffic parameters, as the traffic indicator, that can describe the behavior of short-term, long-term or self-similar traffic patterns. Then, we approximate the cell loss rate in terms of the indicator parameters using a neural network system which consists of a linear combination of a number of sigmoidal functions.