A Domestic Soybean Price Forecasting Model Based On Improved Quantile-RBF Neural Network

Price forecast is the premise of price decision.In order to make scientific decisions according to the trend of the future market price,it is necessary to study the current market and forecast price based on the effective factors.With the continuous development of the market economy,price forecast has become an effective method of production researching,economic planning and so on,which has become a hot topic in recent years.As a large agricultural country,agricultural product market is an extremely important part of Chinese market economic system,and the price is the core.Price information is the signal of the trend of agricultural market,that is to say,whether the price of agricultural product is reasonable or not,it not only reflects the development level of agricultural economy,but also anffects the development of related business.As an important research aspect in forecast field,agricultural products price forecast is theoretically important as well as practically urgent in many scopes.This paper focuses on the agricultural products price prediction,and the study is carried out in the following aspects:1.A forecasting model,which is based on radial basis function(RBF)neural network with the estimation of quantile regression(QR),is proposed in this paper.The prediction model has advantages as follows:it can effectively simulate the nonlinear relationship of'series and make full use of all the information of the data to forecast.2.On the basis of the above framework of forecasting model,the structure and parameter design of the forecasting model based on improved Q-RBF neural network is studied.In order to enhance prediction accuracy and convergence efficiency,the gradient descent method and genetic algorithm(GA)are applied in combination to design the model.At first,the feasibility of the combination of gradient descent method and genetic algorithm is discussed;then the internal structure and parameter of the prediction model is introduced;thirdly,detailed design of the improved algorithm is studied,including the combination algorithm based on gradient descent method and genetic algorithm;at last,the improved algorithm is summed up.3.The domestic soybean price forecasting model based on improved Q-RBF neural network is developed,and the comparison between different models is done.At the same time,we come to conclusions as follows:this forecasting model has obvious advantages in two aspects:prediction accuracy and convergence efficiency.It can be widely used in the field of agricultural price forecast.