本文以潮州市建成区和近期规划区为研究区,采用训练样本获取先验概率建立朴素贝叶斯分类器,以栅格点为单位,将各栅格点的土地定级因素作用分值作为输入变量,利用朴素贝叶斯分类器进行土地定级。在作用分值确定方法上,采用障碍距离代替传统直线距离,以达到客观反映点、线等要素对城镇土地使用价值作用的程度。最后,对顾及障碍物的朴素贝叶斯定级结果分别与空间聚类结果及未顾及障碍物的定级结果进行比较,结果表明本文所提出的方法在土地定级研究中具有一定的优势,能更加真实地反映城镇土地使用价值的空间分布特征。
Traditionally,data mining models and spatial clustering methods are employed in the evaluation of urban land gradation.Given the subjectivity of these methods and their lack of artificial intelligence,intelligence models are required.Here,we propose a na ve Bayesian classifier to determine urban land gradation.Chaozhou in Guangdong province was selected as a case study area.First,according to the suggestions of experts and planners the training area of each gradation was selected.Then,the factors for commercial land gradation in Chaozhou were selected on the basis of the characteristic of the area and the‘Regulations for gradation and classification on urban land'.The function values of factors for commercial land were inputted into the model to determine commercial land gradation.To retrieve the obstacle distance,spatial analysis in AcrGIS was used.The advantage of obstacle distance is that it takes the actual distance in the real world into consideration.For example,if there is a mountain or water body between point a and point b in the real world,the obstacle distance between point a and point b will not pass through the impediment but around the obstacle.Oppositely,the Euclidean distance will pass the mountain or water body directly and is unlike real world.Within this context the impact of factors on urban systems is reflected by obstacle distance.Finally,to confirm the fitness our proposed Bayesian model and obstacle distance,a comparison of the result of the Bayesian based commercial land gradation method considering obstacles with the gradation arising from spatial clustering was carried out.This comparison showed that Bayesian methods are beneficial the research in this area,and accurately reflect the distribution of urban land quality.