Modeling and prediction of foodstuff output based on ε-SVR method
基于ε-SVR的粮食产量预测模型及应用

A new foodstuff output prediction method was investigated based on ε-support vector regression (ε-SVR). The statistical data of foodstuff output of Zhejiang Province in the last 14 years were applied as the analytical matrix. The influence factors of foodstuff output were the following 10 aspects, including agricultural practitioners, cereal sown area, total foodstuff plant area, total power of farm machinery, electricity consumed in rural area, area of affected crops, disaster-affected areas, foodstuff purchasing price of preceding year, effective irrigated area and fertilizing amount. The foodstuff output data of year 1991-2002 were applied as calibration set to develop ε-SVR model, and the prediction precision was 98.47%. The foodstuff output of year 2003 and 2004 was used as validation set, and the prediction precision by the above developed ε-SVR model were 97.5% and 95.8% for year 2003 and 2004, respectively. The results above indicate that ε-SVR is suitable for the analysis and short-term prediction of foodstuff output, and ε-SVR supplies a new method for the prediction of foodstuff output.