径流预测是水文科学研究的重要内容。针对径流时间序列的特性,本文尝试建立了一种惩罚加权支持向量机回归的径流预测模型。通过与BP神经网络和标准支持向量机的结果进行比较,表明该模型预测精度更高,可以用于河川径流的中长期预测。
Runoff forecast is a fundamental part of hydrology.This paper proposes a penalty weighted support vector machine regression model to better describe the features of runoff time series.In comparison to BP neural network and standard support vector machine regression,this model is more accurate and suitable for mid-long term runoff forecast.