Wavelet-SVM-ARMA Combination Forecasting Of Agricultural Product Price

It is a great significance for agricultural product price early warning mechanism to establish accurate forecasting models which are an important tool for the macroscopic readjustment and control of national agricultural market. Because agricultural prices are affected by market supply and demand, production costs, market circulation, natural climate and unexpected event, they show some features such as frequent short term fluctuation, and the overall volatile and unstable. Agricultural prices drop will increase the risk of agricultural producers and result in the reduction of the farmers' income, farmland waste and other adverse effects. Therefore, form macroscopic aspect, the forecasting of agricultural product prices may provide valuable information for the government, while from the microscopic aspect, it is useful for the farmers to avoid the risk of falling prices. It is a great significance for establishing combination forecasting model according to the non-linear, non-stationary and periodic characteristics of agricultural products.This paper first introduces the wavelet theory, support vector machine(SVM) and time series analysis that are three important mathematical tools already applied in the research of agricultural product prices. Then we present a wavelet-basis ARMA-SVM model the forecasting of agricultural product price. The selection of the sample data, the preliminary treatment and model evaluation methods are also described in detail. Based on Chinese pig average prices from 2006 to 2013, we show that the combination forecasting model has better accuracy than single model.