论文部分内容阅读
大豆是重要的经济作物,同时也是我国市场化和国际化程度最高的大宗农产品,对其价格进行预测具有重要意义。采用Q-RBF神经网络模型对国产大豆价格进行预测,该模型具有如下两个特点:(1)通过分位数回归功能来描述大豆在不同价格水平下的分布特征;(2)通过RBF神经网络结构来刻画大豆价格的非线性关系。在模型参数优化时,由于遗传算法是一种全局搜索优化方法,但是搜索速度慢、对初始值具有一定依赖性;而梯度下降法具有收敛快,对初始值没有特定要求等优点,所以本文提出遗传算法与梯度下降法相结合的混合改进算法,其基本思想是利用梯度下降法的局部寻优能力加快遗传算法的收敛速度。采用2010年1月-2015年12月的国产大豆月度价格数据进行预测研究,结果表明,算法收敛速度较快,模型预测精度较高,是可以泛化应用的预测模型。
Soybean is an important cash crop and also a bulk agricultural product with the highest degree of marketization and internationalization in China. Therefore, it is of great significance to forecast its price. The Q-RBF neural network model is used to predict the price of domestic soybean. The model has the following two characteristics: (1) The quantile regression function is used to describe the distribution characteristics of soybeans at different price levels; (2) RBF neural network Structure to characterize the non-linear relationship between soybean prices. When the model parameters are optimized, the genetic algorithm is a kind of global search optimization method, but the search speed is slow and has certain dependence on the initial value. However, the gradient descent method has the advantages of fast convergence and no specific requirements on the initial value. Therefore, Genetic algorithm and gradient descent method combined with improved hybrid algorithm, the basic idea is to use gradient descent method to optimize the local optimization ability to accelerate the convergence rate of genetic algorithms. The forecast data of domestic soybean from January 2010 to December 2015 are used to forecast the results. The results show that the algorithm converges fast and the prediction accuracy of the model is high, which is a prediction model that can be generalized and applied.