| 利用模糊密度聚类和双向缓冲区自动识别热点区 |
| |
| 引用本文： | 崔晓杰,王家耀,巩现勇,赵耀.利用模糊密度聚类和双向缓冲区自动识别热点区[J].武汉大学学报(信息科学版),2019,44(1):84-91. |
| |
| 作者姓名： | 崔晓杰 王家耀 巩现勇 赵耀 |
| |
| 作者单位： | 1.信息工程大学地理空间信息学院, 河南 郑州, 450000 |
| |
| 基金项目： | 中国工程院重点咨询研究项目2017-XZ-13 |
| |
| 摘 要： | 通过数据挖掘手段获取聚集模式（即热点）等地理空间知识是地理信息智能化服务的基础和前提。点群聚集模式的提取本质上是热点及其边界（热点区）的探测。首先分析了使用空间聚类提取热点并以凸壳表达热点轮廓的不足，进而提出一种利用模糊密度聚类和双向缓冲区的热点区自动识别方法。该方法借鉴模糊集理论，通过计算对象之间的模糊隶属度改进基于密度的聚类算法，用以提取点群的聚集模式；在此基础上，将模糊隶属度作为对象间的影响程度，采用正负缓冲区建立热点边界。以郑州市城区的科研机构点为例进行实验，结果表明，提出的方法既能有效区分空间点的类型（噪声点与非噪声点），又能生成连续平滑的热点边界，总体效果优于对比方法。
|
| 关 键 词： | 地理空间知识 热点区识别 空间聚类 模糊隶属度 缓冲区 |
| 收稿时间： | 2018-09-13 |
Hotspot Area Recognition by Using Fuzzy Density Clustering and Bidirectional Buffer |
| |
| Affiliation: | 1.Institute of Geospatial Information, Information Engineering University, Zhengzhou 450000, China2.Research Institute of Henan Spatio-Temporal Big Data Industrial Technology, Zhengzhou 450018, China |
| |
| Abstract: | Obtaining geospatial knowledge such as aggregation mode (i.e. hotspot) by data mining is the basis and premise of geographic information intelligent service. The aggregation mode extraction from point group is the detection of hotspots and their boundaries (hotspot areas) essentially. This paper firstly analyzes the shortcomings of the DBSCAN (density-based spatial clustering of applications with noise) -convex hull method for hotspot area recognition, and then proposes an automatic method of hotspot area generation using fuzzy density clustering and bidirectional buffer. There are two parts in this method:①Based on the theory of fuzzy sets, the fuzzy membership is calculated to improve the DBSCAN; ②The boundaries of hotspots are generated using positive-negative buffer method according to the influence radius calculated by the fuzzy membership formula. The experimental results show that this method can reflect the spatial pattern of the scientific situation. Besides, noises can be distinguished from points, thus ensuring there are no noise points in the hotspot area. Moreover, the hotspot boundaries are not only continuous and flat, also can reflect the actual shape and range of reasonable hotspot areas. Compared with the DBSCAN-convex hull method and the kernel density-contour method, the hotspot area recognized by the method proposed in this paper is better. |
| |
| Keywords: | |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《武汉大学学报(信息科学版)》浏览原始摘要信息 |
|
点击此处可从《武汉大学学报(信息科学版)》下载全文 |
|
