Lasso法在我国农业用水效率模型的应用研究

Study on the Model of Agricultural Water Use Efficiency in China by Lasso Method

在收集全国31省(市、区)1997—2010年农业用水效率及其影响因素相关指标数据基础上,采用Lasso法分区建立以灌溉水利用系数为应变量、影响因子为自变量的农业用水效率分析模型,并筛选出各分区主要影响因子。结果表明,所建模型平均绝对误差都小于6%,表明所建模型精度较高。农民人均收入、人均水资源量、节水灌溉面积、人均耕地面积、降水量为影响农业用水效率的核心因子。

Based on the data of agricultural water use efficiency and related factors from 1997 to 2010 in the 31 provinces （cities or regions）, analysis model of agricultural water use efficiency was established by Lasso, which was used to select the most important factors in different regions. The results indicated that the average absolute errors of the fitting and testing models were all lower than 6 G, and the established model was reliable for forecasting the future efficiency of agricultural water use. Income per farmer each year, water resources per capita, water-saving irrigation area, agricultural acreage per 10 thousands people and precipitation were the key factors influencing agricultural water use efficiency.