Presentation on theme: "Urban Computing with Taxicabs"— Presentation transcript:
1Urban Computing with Taxicabs TMSGCappuccinoOct. 12th, 2011
2About the Authors..Yu Zheng, Yanchi Liu, Jing Yuan, Xing Xie from M$RAYu ZhengUbicomp - Ubiquitious computing groupWSM - Web Search & Mining GroupYu Zheng’s weibo :
3About the Authors..Yu Zheng, Yanchi Liu, Jing Yuan, Xing Xie from M$RAYanchi Liu, also from University of Science and Technology BeijingPaper published by ACM on Sept ( ? )
4Agenda About the authors(finished) Introduction Overview Evaluation SettingsEvaluation ResultsRelated WorksConclusionComments
5Introduction GPS equips are generally used nowadays. GPS-equipped taxicabs can be viewed as ubiquitious mobile sensiors, gathering datas of traffic flows.
6IntroductionThis paper’s goal is to detect the flawed and less effective urban planning in a city according to the GPS trojectories of taxicabs.Saying Beijing has 35 million personal trips per day created by various kinds of vehicles and that 1.44 million personal trips are generated by taxi. Thus, the percentage of taxi trips from total trips are 4.2%. The author believed 4.2% is a significant sample reflecting to urban traffic flow.
7OverviewLet’s introduce the variables one at a time..
17Overview 2. Transition construction Black point represents a region pair, while blue and red points are the projections of these re- gion pairs on XZ and YZ spaces.
18Overview 2. Transition construction Most taxis intend to travel through a shortcut instead of the roundabout route if the shortcut is effective. On the contrary, if most taxis pass additional regions, that means the route directly connecting two places is not very effective.
21OverviewIn this paper, which means there’s no region pair a(p, q) having a lower speed and bigger theta than those belongs to skyline.
22Overview The detected skyline is comprised of 3 kinds of region pairs. 1. Very small E(V) and theta, which means the two regions are connected with some direct routes while the capacity of these routes are not sufficient.
23Overview2. A region pair with a small E(V) and big theta, which means people have to take detours and also suffer from a very slow speed. Worse case. 3. A region pair with a big E(V) and big theta. Meaning that the travel speed is fast but far, still has flaws.
28Overview 2. Mining frequent sub-graph patterns Mining the association rules.The mined association rules can consist of over 2 patterns. E.g. g1, g2 => g3. Also, these association rules may NOT be geospatially close to each other.
30Evaluation Settings Map data Road network of Beijing, consist of 106,579 road vertices and 141,380 road segments.Picked out 25,262 road segments w/ leveling from 0 to 6. Use only the 0 to 2 level. (0 is highest representing highways)Create 444 regions in result.106,579 路口 ; 141,380路段
31Evaluation Settings Verify the detected flaws in the following 2 ways: Verify the urban plannings that had been implemented between the times of the two datasets.Check if some flaws that have been detected in both datasets by our methods embodied in the future urban planning of Beijing.
33Evaluation ResultsSum up that the traffic conditions in Beijing become worse in 2010 than 2009.
34Evaluation Results Taxi drivers took fewer passengers than b4. The average speed dropped between the two years.In short, the traffic condition became worse.
35Evaluation ResultsMost regions becomes shallower in 2010 than 2009, especially in hot areas.The travel speed of taxis in these regions decreased.
36Evaluation ResultsBy looking at the results, we observe two aspects: 1. Some flawed planning occurring in 2009 disappeared in The number of regions having defects increased in 2010 beyond 2009 and some flaws occuring in 2009 still exist.
37Evaluation ResultsSome flawed planning occurring in 2009 disappeared in 2010.
38Evaluation Results Some flawed planning that still exists in 2010. The planning of these two subway lines denotes that the urban planner has recognized the problem existing in the regions, justifying the validity of the results generated using our method.
39Evaluation ResultsCan also found association rules between the detected patterns.Support = 0.05Confidence = 0.7
40Evaluation ResultsWe will show more interesting results and a live demo during the presentation at the conference.
41Related works Mining taxi trajectories Urban computing Effectiveness, like what I’ve presented last time.Destination prediction, like what Danny presented last time.Urban computingMost papers do their researchs on social computing and some user interaction stuff while this paper explore the urban computing from the perspective of urban planning.
42ConclusionThis paper detect the flaws in the existing urban planning of a city using the GPS trajactories of taxis traveling in the urban areas.
43Conclusion Future plans: 1.Studying the geographic features of a region, such as the road segments and POI.2. The purpose of people’s travel.Point of interest
44Comments Easy reading Interesting and well-applied on a useful case Chart arrangement thing again.Like reading a novel or comic, no highlights.