Short-term Price Forecast Model For Fresh Agricultrual Products Based On Price Decomposition

Stabilizing the market price of agricultural products has always been a major event related to national economy and people’s livehood. In order to avoid radical volatility in agricultural products price, enhancing monitoring and early-warning in agricultural market is mentioned several times in the No. 1 Central Document and government reports in recent years. From the view of price formation, the volatility of agricultural products which highlights the short-term market risks and increases the difficulty of price forecast is affected by a variety of factors especially for fresh agricultural products. At present, existing problems of agricultural market have interfered with market economic order and people’s normal life, counting against the development of China’s agriculture. Facing with the constant change of market situation, agricultural producers are hard to grasp valuable information of market supply and demand and price to make reasonable decisions, agricultural administrative departments are unable to provide regulations and control measures in advance either since they have no efficient information for short-term price volatility, consumers are easily aroused much fear by frequent fluctuations because of lacking authoritative information in time. All these problems of insufficient information speed up the vicious spiral of price fluctuation. Therefore, forecasting the short-term price and trend of agricultural price has very important theoretical and practical significance to promote the stability of agricultural production, improve peasants’ income, and guarantee the product supply in agricultural market.Base on modern statistical learning theory, western economics, econometrics, and agricultural information analytics, the core method of this paper is to divide first and integrate after. Using the modern time series analysis and prediction method, empirical mode decomposition method, Hilbert-Huang transform method, support vector regression method, combining model method, and empirical analysis method, the prices of agricultural products are decomposed and the results are analyzed to explore the deeper meaning behind.The object of this paper is fresh agricultural product. First of all, ginger, cabbage, potato, grass crap, and pork are selected as the representative objects by their trading volumes, yield, and particularity. Then, the fluctuations of their prices are briefly analyzed, and the reasonable ranges of fluctuations for each species are found. Next, the HP filter method, Census X12 method, and EMD method are used to decompose each price series, and the advantages and disadvantages of each method are discussed. Moreover, the EMD method and HHT method are applied to analyze those factors influencing agricultural prices. The factors are divided into three categories: events, uncertainties, and trend. Combing event study, the impacts of events are quantified and discussed. By the research, taking influencing period, influencing type, and influencing degree as breakpoint, the influencing factors and influencing results are interpreted.After comparing the forecasting results of the ARIMA and SVR models, SVR is selected to be the method of forecast and combined with the three decomposition methods to forecast future prices for agricultural products. By evaluating the results, EMD-SVR model is proven to be the best forecast model to forecast the price of fresh agricultural products. Meanwhile, by applying the research results, a trinity short-term price monitoring and early-warning mechanism of agricultural products is established based on EMD method from the aspects of forecast results, volatility origin, and historical experience. Policy proposals are also suggested to stabilize the market price and guarantee peasants’ income.