基于多分类支持向量机的网络流量预测方法

 针对支持向量机网络流量预测误差较大的问题，提出一种基于多分类支持向量机的网络流量预测方法。该方法在网络流量数据训练阶段通过数据编码，使多分类支持向量机的输出逼近编码值，在预测阶段通过数据解码，将多分类支持向量机的输出转换为实际的网络流量预测结果，从而有效地降低了预测误差。实验结果显示，该方法的预测结果与实际采集的网络流量数据具有相同的变化趋势；在同等实验条件下，该方法预测结果的均方根误差为0.487，而单一支持向量机方法、BP 神经网络方法预测结果的均方根误差分别为1.0954 和2.3642，表明基于多分类支持向量机的网络流量预测方法具有更高的准确性。

A Network Traffic Prediction Method Based on Multi-class Support Vector Machine

Support vector machines (SVMs) have been used for network traffic prediction, but there often exist large prediction errors. This paper presents a network traffic prediction method based on multi-class support vector machine. Through establishing an encoding method in the training phase, the training output is binary encoded, and is established by one to one correspondence with the training input, training the multi-class support vector machine. By constructing a decoding method in the prediction phase, the prediction output is binary decoded, approaching the real value, realizing network traffic prediction and reducing the prediction errors. In simulation experiments, comparison of the real network traffic data and prediction results shows that they have the same evolution trends. Under the same condition, the accuracy of single SVM algorithm is about two times that of BP algorithm, while the accuracy of the proposed algorithm is about two times that of single SVM algorithm. The experiment results show that the proposed method has higher prediction accuracy in contrast with existing algorithms such as those of single SVM and BP neural network.