A Scalable, Geospatial Segmentation Method for Visiting Trend Analysis

Recently location data has been used for intelligent decision making in various purposes, such as marketing, tourism dynamics analysis and sales. Location intelligence allows users to store, visualize, and analyze user behavior from geospatial data in real time. Such algorithms often rely on identifying users visiting the relevant point of interest (PoI). Visitor statistics of such PoI’s Parking lot has been used to calculate user trends and store dynamics. However, the tagging of such parking lots is often manual and remains a painstaking task. We outline a Floating Car Data based technique that utilizes an unsupervised clustering technique to detect parking lot polygons for a given PoI. We also utilize a novel application of image processing as a noise reduction strategy for the clustering output. Moreover, we take advantage of OpenStreetMaps for end-to-end detection of parking lots polygons. Our approach improves the baseline performance significantly.