The urban facility, one of the most important service providers is usually represented by sets of points in GIS applications using POI (Point of Interest) model associated with certain human social activities. The knowledge about distribution intensity and pattern of facility POIs is of great significance in spatial analysis, including urban planning, business location choosing and social recommendations. Kernel Density Estimation (KDE), an efficient spatial statistics tool for facilitating the processes above, plays an important role in spatial density evaluation, because KDE method considers the decay impact of services and allows the enrichment of the information from a very simple input scatter plot to a smooth output density surface. However, the traditional KDE is mainly based on the Euclidean distance, ignoring the fact that in urban street network the service function of POI is carried out over a network-constrained structure, rather than in a Euclidean continuous space. Aiming at this question, this study proposes a computational method of KDE on a network and adopts a new visualization method by using 3-D "wall" surface. Some real conditional factors are also taken into account in this study, such as traffic capacity, road direction and facility difference. In practical works the proposed method is implemented in real POI data in Shenzhen city, China to depict the distribution characteristic of services under impacts of multi-factors.
Network Kernel Density; Network Analysis; POI (Point of Interest); Spatial Statistics
