Zufiria Pedro J, Hernández-Medina Miguel Á
ETS Ingenieros de Telecomunicación, Information Processing and Telecommunications Center (IPTC), Universidad Politécnica de Madrid, 28040 Madrid, Spain.
Entropy (Basel). 2019 Aug 8;21(8):774. doi: 10.3390/e21080774.
Based on a sample of geolocated elements, each of them labeled with a (not necessarily ordered) categorical feature, several indexes for assessing the relationship between the geolocation variables (latitude and longitude) and the categorical variable are evaluated. Among these indexes, a new one based on a Voronoi tessellation presents several advantages since it does not require a variable transformation or a previous discretization; in addition, simulations show that this index is considerably robust when compared with the previously known ones. Finally, the use of the presented indexes is also illustrated by analyzing the geolocation of communities in some communication networks derived from Call Detail Records.
基于一组地理定位元素样本，每个元素都标记有一个（不一定有序的）分类特征，评估了几个用于评估地理定位变量（纬度和经度）与分类变量之间关系的指标。在这些指标中，基于Voronoi镶嵌的一个新指标具有几个优点，因为它不需要变量变换或预先离散化；此外，模拟表明，与先前已知的指标相比，该指标具有相当强的稳健性。最后，通过分析一些从通话详单记录导出的通信网络中社区的地理位置，说明了所提出指标的使用情况。
