A Study on the Forecast of the Price of Agro-products
Since China is a large agriculture country, it is significant to establish agro-product price data market for predicting the price trend in short, median and long term, in order to guide the agriculture production. Therefore, we selected pork as an example, and collected the price information of the pig to be slaughtered, corn, bean pulp, and so on, from 14 sampling sites in Fujian province since 2008, and established a agro-product price data market for analysis. First, we used gray association model for clustering, then constructed the prediction model using algorithms of ARMA, multi-variable linear regression and neural network. The system automatically constructed models for every possible parameter combinations in the range of parameters predefined by experts, and selected two prediction models with the minimum extrapolating errors for specific products. The predicted value of the product was derived from the weighted average of the values produced by these two models. In the two-years tracking analysis, we compared 1358 predicted values and real ones, and computed the relative forecasting errors, the results showed that 74.54％ of samples had an error lower than 1％, while 95.4％ had an error less than 5％, therefore, the model had a relative good prediction effect, and could be used to guide agriculture production. Meanwhile, the system also conducted data mining analysis for the price range of pork, and quantified the agro-product consumption motivations, providing support for government in decision-making.
agro-product price forecasting data mining
Zhang Jinyi Chen Shao-dong Chen Ke
Agriculture Bureau ofFujian Province. Fuzhou 350003, P.R.China Fujian Soil Conservation & Agricultural Development Asian Development Bank Loan Project Management O Harbin Institute of Technology. Harbin 150001, P.R.China
国际会议
2010 International Symposium on Agricultural Ontology Service(农业本体服务2010年国际学术研讨会)
北京
英文
146-152
2010-10-30（万方平台首次上网日期，不代表论文的发表时间）