一种基于重排检验的时空聚类方法

发布时间：2018-07-06 06:43

本文选题：时空数据挖掘 + 分布模式　；参考：《武汉大学学报(信息科学版)》2017年04期

【摘要】：融合时空邻近与专题属性相似的时空聚类是挖掘地理现象时空演化规律的重要手段。现有方法需要的聚类参数许多难以获取,影响了聚类方法的可操作性与聚类结果的可靠性。提出一种基于重排检验的时空聚类方法。首先,通过重排检验发现时空数据集中的均质子区域;进而,采用均方误差准则合并均质子区域内的时空实体生成时空簇,并通过簇内重排检验自动识别聚类合并的终止条件;最后,借助时空拓扑关系在保证结果精度的前提下发展一种快速重排检验的方法,提高了聚类方法的运行效率。通过实验和比较发现,该方法一方面可以发现不同形状、大小的时空簇,聚类质量优于经典的ST-DBSCAN方法;另一方面聚类过程中人为设置参数的主观性显著降低,提高了聚类方法的可操作性。
[Abstract]:Spatiotemporal clustering, which is similar to thematic attributes, is an important means to excavate the temporal and spatial evolution laws of geographical phenomena. Many of the clustering parameters needed by the existing methods are difficult to obtain, which affects the maneuverability of the clustering methods and the reliability of the clustering results. A spatiotemporal clustering method based on rearrangement test is proposed. Firstly, the average proton region in the spatio-temporal data set is found by rearrangement test, and then the spatio-temporal cluster is generated by combining the space-time entities in the mean square error criterion, and the termination conditions of clustering merging are automatically identified by intra-cluster rearrangement test. Finally, with the help of spatio-temporal topological relationship, a fast rearrangement test method is developed to improve the running efficiency of the clustering method on the premise of ensuring the accuracy of the results. Through experiments and comparisons, it is found that on the one hand, the clustering quality of this method is superior to that of the classical ST-DBSCAN method, and on the other hand, the subjectivity of the artificial setting parameters is significantly reduced in the clustering process, and the clustering quality is better than that of the classical ST-DBSCAN method. The maneuverability of clustering method is improved.
【作者单位】：中南大学地球科学与信息物理学院;日本电气中国研究院;
【基金】：国家自然科学基金(41471385,41171351,41601410) 资源与环境信息系统国家重点实验室开放基金中南大学中央高校基本科研业务费专项资金(2016zzts084)~~
【分类号】：P208

【相似文献】