Learning locality maps from noisy geospatial labels

Abstract

E-commerce and logistics operations produce a vast amount of geospatial data labelled with postal addresses. The data has great potential to mine geospatial knowledge, and we demonstrate that regional maps can be automatically built using the same. We propose an algorithm to construct non-overlapping polygons of the localities at a city level. The algorithm involves non-parametric spatial probability modelling of the localities followed by locality classification of the cells in a hexagonal grid. We show that our algorithm is capable of handling noise, which is significantly high in our setting due to the small scale of localities. A property about the noise and the correct information is presented such that our algorithm infers a correct locality polygon. We quantitatively measure the accuracy of our system by comparing its output with the available ground truth. We also discuss multiple applications of the generated maps in the context of e-commerce and logistics operations.