Abstract: Associated with Particle Swarm Optimization (PSO) and Least Squares Support Vector Machine (LSSVR), this study proposes a soil fertility evaluation model based on PSO-LSSVR. Ten soil indicators were selected as elevation indexes, which included organic matters, total nitrogen, available phosphorus, available potassium, cation exchange, capacity, acidity-alkalinity, volume weight, clay dispersion, water stable aggregated and dispersion ratio. The black soil in Jilin province was used example to construct soil fertility evaluation model. The result was compared with two other methods (extension-element method and ordinary SVM method), the soil fertility levels of most samples were consistent, sample 2 and sample 13 were classified as grade Ⅳ and grade Ⅲ by PSO-LSSVR model respectively, which were in line with the actual situation, which showed that PSO-LSSVR model was a suitable soil fertility evaluation model and the result accurately reflected soil properties.

Key words: least square support vector machine regression, evaluation, particle swarm optimization, soil fertility

摘要： 通过引入粒子群算法(PSO)和最小二乘支持向量机(LSSVR)，提出基于PSO-LSSVR的土壤肥力评价模型。选取有机质、全氮速效磷、速效钾、阳离子交换量、酸碱度、容重、黏粒、水稳性团聚体和分散率等10种评价指标，以吉林省黑地为例，建立土壤肥力评价模型。同时与物元可拓法、普通SVM模型的评价结果进行比较；3种方法的多数样本评价结果基本一致，对于样本2、样本13，PSO-LSSVR模型分别定为Ⅳ级、Ⅲ级，符合实际情况；表明PSO-LSSVR是一种适用且能准确反映土壤特性的土壤肥力评价模型。

关键词: 最小二乘支持向量机, 评价, 粒子群算法, 土壤肥力