Research On Fruit Price Forecasting And Fluctuation Early-warning

Fruit is the fourth largest crop in China. In 2013, China’s fruit production was 250.93 million tons, and achieved 696.9 billion yuan in output value. However,China’s per capita consumption of fresh fruit is only 37.8kg, far below that of those developed countries, which means the fruit consumer market in China has a great potential.But the development of the fruit industry faces many challenges, among which the fruit price fluctuation is typical. It directly affects the enthusiasm of the fruit business operator, and reduces the fruit enterprises to great risks. Fruit price forecasting and volatility early warning are important for fruit businesses.Therefore, the fruit price fluctuation is selected for the research in this thesis. The main contents and conclusions are as follows:First, fruit price prediction is conducted. In this thesis, BP neural network, SVM and ARMA model are used for the annual and monthly fruit price forecast. In the annual forecast, the prediction model is constructed through the index system of supply and demand of fruit market. Prediction error distribution within 5% and 10% by SVM is better than the BP neural network. In the monthly forecast, the prediction model is constructed through time series of the fruit price, in which, prediction error by BP neural network and ARMA for the three fruits is respectively within 5%, while that by SVM is within 1%. By comparison, SVM is nailed down as the forecast and early warning model.Second, fruit price fluctuation’s early-warning is conducted. Based on forecast analysis, SVM is chosen as the method for fruit price fluctuation warning. This thesis takes the fruit price fluctuation as alarm indicator, and the degree of warning and alarm limits will be determined by the mean and standard deviation of statistical methods. Then, based on the fruit price forecast, the thesis achieved annual and monthly fruit price volatility warning by SVM. The accuracy rate of annual warning by SVM for three fruits is respectively 64.70%, 100% and 94.12%, and that of monthly early warning is respectively 100%, 91.67 % and 83.33%.Third, the study analyzed how the fruit business carry out early-warning response in the context of the fruit price fluctuation. This thesis suggests that the wholesale fruit business should strengthen its own informatization construction, improve fruit price fluctuation forecast and precaution against different warning situations, and optimize the fruit price strategies in accordance with the price fluctuation; thereby they can enhance their capacity to respond to the risk of Price fluctuation.Innovations of this thesis:(1) Quantitative analysis is applied to the macro industrial environment analysis, which enhances the pertinence and effectiveness of management measures.(2) BP neural network is improved through optimal mean value by repeated calculations,which improved the stability and prediction accuracy of the model;(3)Besides, the model selection is improved through comparison among the BP neural network, SVM and ARMA,which optimizes the selection process for forecast and precaution model;(4)The set-up of early warning limit and alert degree is improved. This thesis counts the positive and negative fluctuations respectively, and sets up the early warning limit and alert degree through the separate mean, standard deviation and one-sided confidence interval. It enhances the feasibility and scientificity of the early warning limit and alert degree.