基于最近邻聚类学习算法的模糊业务量预测器  

A Fuzzy Traffic Predictor Using Nearest Neighborhood Clustering Learning Algorithm

作　　者：庞其祥[1] 程时端[1] 

机构地区：[1]北京邮电大学程控交换与通信网国家实验室,北京100876

出　　处：《电子学报》1999年第10期12-14,共3页Acta Electronica Sinica

基　　金：国家"863"项目资助

摘　　要：本文根据模糊逻辑系统中的万能逼近定理［１］ ，提出一个采用最近邻聚类学习算法的模糊业务量逼近器／预测器．通过对自回归（ＡＲ） 过程及ＭＰＥＧ 业务量的预测，显示了该逼近器／ 预测器的准确性和对业务量特性的适应性．This paper,based on the Universal Approximation Theory of Fuzzy system,presents a fuzzy traffic predictor using nearest neighborhood clustering learning algorithm.Its prediction effect is examined through autoregressive (AR) process and real video traffic.The results show that the approximator/predictor is accurate and flexible.At last,the applications of the predictor in dynamic bandwidth allocation is discussed.

关 键 词：ATM 业务量控制 业务量预测 模糊逻辑系统 

分 类 号：TN913.24[电子电信—通信与信息系统]