基于聚类和支持向量机的话务量预测模型

Traffic Forecasting Model Based on Clustering and Support Vector Machine

针对利用单因素时间序列模型进行话务量预测的不足,建立基于模糊C均值(FCM)聚类和支持向量机(SVM)的多元回归话务量预测模型。模型使用FCM算法对话务量的原始样本集聚类,选择与待预测样本特征最相似的样本子集作为训练集,使用SVM训练样本,通过决策回归函数预测话务量。实际话务量数据验证表明,该方法较周期时间序列和神经网络预测方法具有更高的预测精度和泛化能力。

A novel multiple regression traffic forecasting model based on the fuzzy C-means （ FCM ） clustering and the support vector machine （SVM） is presented, to handle the insufficiency brought by the single-factor time series model for telephone traffic forecasting. FCM is used to cluster the original sample set, so that the subset, whose character is the most similar to the sample set to be forecasted, is chosen from the sample set. Then by applying SVM, the decision function is given, to forecast telephone traffic. The verification on the model with real telephone traffic data shows that, comparing with cycle time series and neural network forecast methods, the model performs higher forecasting accuracy and better generalization ability.