Towards Mobilitybased Clustering Siyuan Liu Yunhuai Liu Lionel
Towards Mobility-based Clustering Siyuan Liu*#, Yunhuai Liu*, Lionel M. Ni*# +, Jianping Fan#, Minglu Li+ *Hong Kong University of Science and Technology #Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences + Shanghai Jiao Tong University July 27 th, 2010@SIGKDD 2010
Outline Introduction Related work Mobility based clustering Field study evaluation Conclusion
Outline Introduction Related work Mobility based clustering Field study evaluation Conclusion
Smart City [1] China's urbanization Massive issues and problems City monitoring and management Pervasive information and knowledge Digital technology Data collection, storage and mining Real life data sets Vehicle GPS data sets (one year, two cities) Mobile phone networks data sets (one year, two cities) Hot spot detection in the city [1] Smart City Research Group. http: //www. cse. ust. hk/scrg
Motivation Crowded spots and areas in the city Traffic congestion Event detection Commerci al promotion Vehicle instant locations of sample taxis at 13: 00 PM on 12 th Dec, 2006
Data set Ideal case: we should have all information of all vehicles in the city Reality: only a sample set of all vehicles Taxi GPS data (ID, location, speed, time, direction, status) 0. 3% of the two million vehicles in Shanghai Could we utilize such a very limited sample set to detect hot spot in the city?
Challenges Dense? Sparse! Extremely limited sample set Sparse? Dense! Notable location error Could density based clustering handle it?
Methodology Observation The low speed may indicate that the area is crowded Method Mobility-based clustering Study the speed (mobility) instead of the density Moving objects as sensors
Outline Introduction Related work Mobility based clustering Field study evaluation Conclusion
Related Work Density based clustering • Partitioning methods • Hierarchical methods • Density based methods • Grid based methods • Model based methods etc. • Raw data based methods • Feature based methods • Model based methods etc. ØWhat if the mobility is high? ØWhat if the density is poor? ØWhat if the location is lossy? 10
Outline Introduction Related work Mobility based clustering Field study evaluation Conclusion
Mobility-based Clustering Roadmap 12
Object Mobility Model Speed estimation Road network grid Interpolation Direction distinguishing Speed spectrum of road direction Speed spectrum of reverse direction Nan. Pu Bridge 13
Spot Crowdedness Model Linear crowdedness function Statistical crowdedness function 14
Crowdedness Model Validation 1. Taxis traces 2. Buses traces 15
Learning in Practice Learning in Mobility Based Clustering Characterizing spots α, г Hot spots and hot regions Spot crowdedness Sensor object profiling Temporal hot spots Evolutionary hot regions 16
Learning in Practice Characterizing spots 17
Learning in Practice Sensor object profiling 18
Learning in Practice Sensor object profiling 19
Learning in Practice Hot spots and hot regions NOT hot spot, even dense sample points Hot spot, even sparse sample points 20
Learning in Practice Temporal hot spots Event detection Temporal consistence
Learning in Practice Evolutionary hot regions Area difference ratio Crowdedness difference ratio 22
Outline Introduction Related work Mobility based clustering Field study evaluation Conclusion
Field Study Evaluation 24
Field Study Evaluation 25
Outline Introduction Related work Mobility based clustering Field study evaluation Conclusion
Conclusion and Future Work Contributions Mobility-based clustering model Key factors on spot crowdedness Hot spots and hot regions Future work More accurate speed information More accurate location information
Thanks for your attention!
- Slides: 28