尾矿坝浸润线序列的支持向量机预测研究

Prediction of Infiltration Route Series in Tailing Dam by the Support Vector Machine

从尾矿库监测系统功能出发,结合尾矿库安全监测数据内容和特点,在充分认识数据挖掘技术如何处理和应用的基础上,提出一种基于单因素时间序列支持向量回归机的浸润线预测方法。从尾矿库监测数据中选取有效样本,运用留一法对支持向量回归机参数进行优化,建立预测模型。结果表明:该方法能在小样本、高精度要求下对浸润线进行准确预测。

In view of the functions of monitoring system of tailing dam,combined with the content and features of safety monitoring data on tailing dam,and based on full understanding on the data mining techniques and its application,a forecasting method of infiltration route based on univariate time series of support vector regression is proposed.That is to select valid samples from the monitoring data of the tailings dam,and optimize the kernel parameters for SVR（Support Vector Regression） by using the leave-one-out method to establish a forecasting model.The results show that a higher accurate forecasting on the infiltration route can be obtained with few samples by this method.