应用支持向量机方法对北京平原粮田区土壤养分肥力的评价研究

Evaluation of Grain Field Soil Nutrient Fertility in Beijing Plan by Support Vector Machines

将支持向量机方法(Support Vector Machines,SVM)应用于北京平原地区粮田土壤养分肥力评价,与判别分析评价的结果相比较,支持向量机能够较好的揭示研究区域土壤养分肥力的现状。以专家评判为准,判别准确率可达84.62%;其中中等养分肥力以上的样本占58.19%,表明北京平原地区粮田土壤养分肥力的总体水平处于中等偏上。根据支持向量机评价的结果进行Kriging最优内插,并绘制出土壤养分肥力等级图,对各区县肥力情况进行分析,表明房山、海淀、丰台和门头沟属较高肥力区县,怀柔和平谷处于中等水平,其他区县养分肥力水平则较低。

Soil nutrient fertility of grain field in Beijing plain area was evaluated by support vector machines（SVM）. Compared to the result evaluated by discriminant analysis, the result evaluated by SVM was more exact contrast to expert judgment, and the accuracy could be up to 85.95%. According to the result evaluated by SVM, the percentage of sample above the middle level was 58.19%, which indicated the soil nutrient fertility in this area was above the middle level. Using the group map of soil nutrient fertility with Kriging interpolation, a preliminary analysis for all districts was made, which showsed that Fangshan District, Haidian District, Fengtai District, Mentougou District were in high level, and Huairou County, Pinggu County were in the middle, others were in low.