针对当前松毛虫滞后阶数确定方法存在局部最优、耗时长等问题,提出一种基于地统计学(GS)快速定阶的松毛虫发生面积组合预测模型(GS-ARIMA-SVM).首先采用差分自回归移动平均(ARIMA)对松毛虫发生面积进行线性建模预测,然后采用GS对松毛虫发生面积非线性部分进行快速定阶和样本重构,最后采用支持向量机(SVM)对非线性部分进行建模预测,从而获得组合模型预测值.并对辽宁省朝阳市松毛虫发生面积数据进行了仿真实验.仿真结果表明,GS-ARIMA-SVM预测精度明显优于参比模型,更能反映松毛虫发生的复杂动态变化规律.%This paper proposed a dendrolimus punctatus occurrence area combination forecasting model ( GS-ARIMA-SVM) based on geostatistics(GS) for solving local optimal and time consuming long in the traditional optimal delay order method. Firstly, forecasted the linear change discipline of the dendrolimus punctatus occurrence areas by autoregressive integrating moving average ( ARIMA) , and then determined the nonlinear change discipline order by GS, reconstructed the data, predicted the nonlinear part by support vector machine ( SVM) and got the combination model' s forecasting results lastly. It test the proposed model performance by dendrolimus punctatus occurrence area of the Chaoyang city of Liaoning province. The results show that the proposed model improves the forecasting accuracy compared with other models, it can reflect dynamic discipline of dendrolimus punctatus occurrence.
展开▼