用近邻算法预测通信量时间序列

Apply Nearest Neighbor Algorithm to Traffic Time Series Prediction

为了对通信系统进行有效的调控,需要对通信量进行预测,而通信量具有在不同日期遵循不同规律的特点。本文采用基于实例的近邻算法进行时间序列预测,并在考虑动态长度序列、序列特征提取和近似样例的选取上做出改进,取得很好的效果。将近邻预测算法应用到广东省电话网智能管理系统(GTNIMS)中,能够为路由求解提供快速、准确的预测话务量,为更精确的求解创造了条件。

In order to coofigure communication system efficiently, it is necessary to predict the traffic, which has the characteristic of following various rules in different date. This paper proposes time series prediction by Case-based nearest neighbor algorithm and make improvements on considering dynamic length of time series, extracting charac- teristic from time series and selecting nearest case, with which ideal results can be achieved. Being applied to Guang- dong Telecommunications Network Intelligent Management System (GTNIMS), the nearest neighbor algorithm can provide fast, accurate prediction traffic for routing solution and set the stage for more precise solution.