论文部分内容阅读
我国大豆价格受国内外多种因素共同影响,具有非线性、随机性和高噪音等特点,采用传统数学模型进行预测,不仅分析难度大,预测误差也很大。RBF神经网络以其优良的逼近性能而被广泛应用于非线性时间序列预测之中。本文提出一种基于遗传算法优化RBF神经网络的我国大豆价格预测模型,该模型为多维输入单维输出的多变量预测模型,模型的初始输入由大豆价格的历史数据和相关影响因素数据组成。采用遗传算法对RBF神经网络输入层节点数、基函数中心、扩展常数和输出层权值进行优化,模型可以从初始输入变量中自主选择最合适的输入变量组合作为模型的输入。采用2009-2014年的大豆价格数据进行预测研究,用2009-2013年的数据作为训练集,2014年的数据作为测试集,改进RBF神经网络通过自主识别和选取中国大豆进口量、中国消费者信心指数和进口大豆到港分销价格3个因素作为相关影响因素的输入。结果表明:模型预测精度较高、泛化能力较强,能够很好捕捉大豆价格变化规律,可为大豆市场价格的准确预报提供参考借鉴。
The price of soybean in our country is affected by many factors at home and abroad, such as nonlinearity, randomness and high noise. The traditional mathematical model is not only difficult to analyze, but also has great forecasting error. RBF neural network is widely used in nonlinear time series prediction because of its excellent approximation performance. This paper presents a prediction model of soybean in China based on genetic algorithm to optimize RBF neural network. The model is a multivariate prediction model of multi-dimensional input and single-dimensional output. The initial input of soybean model is composed of historical data of soybean price and related influencing factors. Genetic algorithm is used to optimize the input nodes, RBFs, expansion constants and output layer weights of RBF neural network. The model can select the most suitable combination of input variables from the initial input variables as the input of the model. Using 2009-2014 soybean price data for forecasting research, 2009-2013 data as a training set and 2014 data as a test set to improve the RBF neural network by independently identifying and selecting China’s soybean imports, the Chinese consumer confidence Index and imported soybean to Hong Kong distribution price of 3 factors as the input of relevant factors. The results show that the prediction accuracy of the model is high and its generalization ability is strong. It can capture the variation of soybean price well and provide reference for the accurate forecast of soybean market price.