1. Visually Analyzing Latent Accessibility Clusters of Urban POIs
Farah Kamw, Shamal AL-Dohuki, Ye Zhao Kent State University
Jing Yang University of North Carolina Charlotte
Xinyue Ye New Jersey Institute of Technology
Wei Chen Zhejiang University

2. Urban Structures Data Urban structures: Defining urban space
Road Network
Roads are categorized into primary, secondary, residential, …
POIs
POIs are categorized into two hierarchical levels.

3. Urban Mobility Data
Urban trajectories: recording movement behavior traces for humans, vehicles, fleets, public transits, …

4. Accessibility of Urban POIs …
Discover and analyze urban objects for their disparities in the access to essential facilities and services.
Urban POIs can reach different sets of facilities and services depending on a variety of factors.
For example, given a 5 minutes of driving, an apartment complex can access 20 restaurants in Saturday noon, but can only access 4 restaurants in Wednesday noon.

5. Contribution
A new interactive visualization system for discovering hidden urban sub-regions with different accessibility patterns.
Computing Latent Accessibility Clusters (LACs) of the POIs.
A graph-based clustering method to form the LACs.
A Voronoi-based drawing method to visualize the LACs as spatial regions on the map.

6. Visual System Framework
Red box: Pre-processing of POI accessibility
Blue box: Interactive discovery and visualization of LACs

7. Accessibility Function
The accessibility feature of one POI p, is the function of several factors:
Origin POI category 𝛹: hotels
Destination POI category 𝛺: restaurants
Travel time threshold 𝜂: 5 minutes
Current time t: 6pm
Transportation mode 𝜁: by car
Thus, the accessibility function 𝛂 is to compute the number of destination POIs accessible from p as
𝛂(p, 𝛺, 𝜂, t, 𝜁) = ∑ ∀ q ∈ 𝛺 (T(p, q, t, 𝜁) < η?1 : 0)
Here q is one destination POI in 𝛺.
T(p, q, t, 𝜁) computes the travel time from p to q at t with the transportation mode 𝜁. If the travel time is in the threshold 𝜂 then q is reachable from p given these constraints.

8. Spatial LAC Generation
Neighboring POIs with similar 𝛂 values form a Latent Accessibility Cluster (LAC)
LAC generation method include several steps:
Building dual graph from road network
Constructing POIGraph
Computing accessibility functions
Defining accessibility difference/similarity
Cutting graph into subgraph clusters
Generating LACs

9. Spatial LAC Generation Example
Illustrating LAC generation. (a) Creating POIGraph from road network and POIs; (b) Computing accessibility function; (c) Computing accessibility difference; (d) Forming subgraphs that represent the LACs.

10. LAC Visualization
LACs are group of clusters each representing a sub-region.
Their boundaries are visualized on the map within two steps:
Voronoi subdivision: partitions the 2D space into cells by using the locations of city POIs as Voronoi seeds.
Voronoi cell merging: merges two neighboring Voronoi cells into one, if the two seed POIs belong to the same LAC.

11. Data and Case Study Discover LACs in downtown of Hangzhou city, China.
247,642 POIs grouped into 18 categories including real estate, shopping, education, etc.
Traffic information were computed from one month taxi trajectory data in the city.

12. Case Study: Clustering hotels based on their accessibility to tourist attraction places within 5 minutes of driving on Sundays at 10am-12pm.
(A) Map view; (B) LAC details; (C) Chart of reachable POIs.
Five LACs are detected with different numbers of hotels
blue:64, red:49, green:23, yellow:1, pink:1
They can reach different number of tourist attractions
blue:18, red:89, green:166, yellow: 64 and pink: 66.

13. Case Study (cont.)
(a)
(b)
(c)
(a) Reachable tourist attraction places from the blue LAC (b) Reachable tourist attraction places from the green LAC which are far more than the blue one; (c) Traffic speed on roads.

14. Summary
We present a new visualization system to identify and study urban sub-regions of different accessibility patterns.
Useful for urban planning, design, and management
To be improved by:
Integrating public transit
Using real time traffic data
Considering walking and biking modes

15. Thanks
Farah Kamw:
Ye Zhao: