基于SVM的农机装备水平组合预测模型研究

采用支持向量机的组合预测方法,对黑龙江垦区农机装备水平进行预测。在确定单一预测模型的基础上,运用自组织神经网络方法,将权系数确定问题转化为粗糙集理论中属性重要性评价的问题;计算各单一预测方法对组合模型的依赖度、重要度和权系数;利用建立的基于支持向量机非线性农机装备水平组合预测模型,对黑龙江垦区2002—2012年农机装备水平的历史数据进行检验。误差分析表明:该模型对农机总动力、大中型拖拉机、小型拖拉机、大中型拖拉机配套机具和小型拖拉机配套机具的预测平均相对误差为0.471%、1.328%、3.738%、1.193%、3.574%,均低于各单一预测模型的平均相对误差;利用该模型对黑龙江垦区农机装备水平进行预测,到2020年拥有农机总动力999.33万kW、大中型拖拉机88 921台、小型拖拉机38 453台,大中型拖拉机与配套农机具台数比为1.51∶1,小型拖拉机与配套农机具台数比为1.68∶1。所建模型适用于黑龙江垦区农机装备水平的预测。

Research on combined forecasting model for the level of agricultural machinery equipment based on SVM

Combined forecasting method based on support vector machine was adopted to predict the level of agricultural machinery equipment in Heilongjiang Province.On the basis of determining single prediction models,self-organizing neural network method was used to transform the problem of weights determine into a problem of importance evaluation in rough set theory and calculate the degree of dependence,importance and weight coefficient of single forecasting model on combined forecasting model.Based on support vector,non-linear combined forecasting model was established to examine the historical data of Heilongjiang Reclamation Area from 2002 to 2012.Error analysis indicated that the model predicted average relative error for total power of agricultural machinery,large and medium-sized tractor,mini-tractors,large and medium-sized tractors implementation equipment and mini-tractors implementation equipment,which were 0.471%,1.328%,3.738%,1.193% and 3.574% respectively.All values were lower than the average relative error of single forecasting model.This model was adopted to forecast the equipment level of agricultural machinery in the Heilongjiang Province.To 2020,Heilongjiang Reclamation Area will own agricultural machinery total power with 9.9933 million kW,88921 sets of large and medium-sized tractor,38453 sets of small tractors.The ratio of large and medium-sized tractor and its supporting tools will be 1.51:1.The ratio of small tractor and its supporting tools will be 1.68:1.The results showed that the combined forecasting model is suitable for predicting agricultural machinery equipment level in Heilongjiang Province.