本文运用人工神经网络（ANN）的理论与方法,构建了应用于城镇土地定级的SOFM网络模型,将＂聚类＂思想应用于定级评价中,克服了传统方法因子权重过于主观的缺点。本文以深圳市宝安区为例进行了方法验证,并与传统方法进行了比较,结果表明,基于SOFM的定级模型在空间聚合性和准确性方面优化了定级结果,进一步表明了此类人工智能类方法在评价领域的适应性。“,”In this study,the theory and method of Artificial Neural Network（ANN） is used to construct the SOFM network model applied in urban land grading.The idea of＂clustering＂is applied in the grading process,and the defect of subjective factor weighting in traditional method is overcome.Data collected from Bao＇an district,Shenzhen city is used to do method validation and the final result is compared with that of traditional method.It is indicated that the urban land grading model based on SOFM performs better in spatial cluster and accuracy than traditional approach,and further that the Artificial Intelligence method is feasible in the field of evaluation.