基于GSM和SVM的区域年用水量回归预测模型研究

A Regression Model for Forecasting Regional Annual Water-consumed Quantity Based on GSM and SVM

区域年用水量受众多因素影响,具有非线性特点,而且还存在记录时间短、历史数据少等问题。基于支持向量机(SVM)小样本、非线性和泛化能力强的特性,建立了年用水量回归预测模型,利用网格搜索法(GSM)优化参数,并进行精度的检验。将模型应用于民勤县年用水量预测,结果表明:该预测模型的绝对误差和相对误差较小,精度较高,用于该县的年用水量预测是行之有效的。

Regional annual water-consumed quantity is affected by a lot of factors and has nonlinear characteristic.There are shorter recording time,less historical data problems.we built a regression model for forecasting annual water-consumed quantity based on the small sample,nonlinear and generalization ability characteristics of support vector machine,and a grid search method was applied to optimize the parameters,and then the model＇s precision was tested.The model was used to forecast annual water-consumed quantity in Minqin.The result showed that the absolute error and the relative error were small,and the precision was high.It is feasible to forecast annual water-consumption quantity of the county.