基于随机森林模型的生猪价格预测及调控机制研究

Research on prediction and regulation mechanism of pig price based on random forest model

生猪产业在国民经济中的重要地位及自身的“弱质性”决定了国家对生猪产业发展的原则是“市场主导、政府调控”,而国家对生猪产业宏观调控的有效性取决于用科学的方法对生猪价格进行预测。研究得出:随机森林价格预测模型中的两个最优参数ntree、mtry的值分别为450和4,随机森林模型预测的残差平方和0.72,拟合优度为98.25%,说明拟合效果较好;在特征选取上,综合平均下降精度和平均下降基尼系数的排序结果,得出影响生猪价格因素最重要的两个变量是去皮带骨猪肉和仔猪价格;随机森林模型和BP神经网络模型对价格的预测对比发现:随机森林模型中均方误差和平均绝对误差分别为0.0231和0.408,BP神经网络模型中均方误差和平均绝对误差为0.0828和0.71,随机森林在生猪价格预测方面预测精度更优。

The important position of the pig industry in the national economy and its own“weakness”determine that the principle of the country’s development of the pig industry is“market orientation and government regulation”,and the effectiveness of the country’s macro-control of the pig industry depends on the scientific methods to predict the pig price.The results show that the values of the two optimal parameters,ntree and mtry in the random forest price prediction model are 450 and 4,respectively,the sum of residual squares predicted by the random forest model is 0.72,and the goodness of fit is 98.25%,indicating a good fitting effect;in feature selection,based on the sequencing results of average decline precision and average decline Gini coefficient,it is concluded that the two most important variables affecting the pig price are the price of peeled bone pork and piglet;the comparison of price prediction between random forest model and BP neural network model shows that the mean square error and mean absolute deviation in the random forest model are 0.0231 and 0.408,respectively,and the mean square error and mean absolute deviation in BP neural network model are 0.0828 and 0.71.The prediction accuracy of random forest in pig price prediction is better.