Smarter Transportation through on Big Data - Best Practices of Seoul Metropolitan Gov. - Oct. 8. 2015. Kim, Ki-Byoung Chief Data Officer / Director Data & Statistics Division Seoul Metropolitan Government 0 / 29 e-mail: pskbkim@gmail.com
Contents Before starting Introduction & background Late night bus story of Seoul Taxi analysis What SMG has learned Plan & direction 1 / 29
https://www.youtube.com/watch?v=nv-qeq-6yow 2 / 29
Traditional, smart city vs. Smarter city through data Traditional City Smarter City City Urban activities Urban activities Smart Big platform data platform Optimization New value creation City infrastructure Urban infrastructure 3 / 29
What we have been doing so far Area Topic Period Objectives (Expected) Results Transport Safety Transport Night bus route optimization Traffic accident analysis for minorities `13.1H `14.2H Transport Taxi analysis `14.2H Welfare Welfare Location analysis of Life double cropping center Location analysis of Senior welfare center `14.1H `14.1H Welfare Disabled taxi analysis `14.2H Admin City PR booth analysis `14.1H Admin Tourism Location analysis for ATM of civil service Flow analysis of foreign tourists `14.2H `14.1H Route optimization Prepare policies to reduce traffic accidents of children and senior citizen Provide more chance to catch a taxi during mid night Select the best location of life double-cropping center Select the best location of senior leisure centers Reduce waiting time of disabled taxi Select best location of public PR Select prioritized ATM locations Promote foreign tourism with better service Night bus coverage > 42% with 30 buses (daytime buses >7,000) Number of accidents will become 50% in 3 years Provide 5% more chance to catch a taxi in midnight Every facility will provide best coverage for the citizen Every facility will provide best coverage for the citizen Reduce waiting time by 10% More PR s with same booths Provide more ATM services with planned number of devices Increase foreign tourist by 10% 4 / 29
Case 1: Late-night bus routes Why Late-night bus? Response of the City Buses don t run by the time I get off work. I don t have a car. I hope there will be buses available at late night..!! @gu**** No public transportation in 01:00 AM ~ 05:00 AM Let s set-up Late night bus routes Facing Problems 1.Limited resources bus, drivers, budget 2. Where is traffic demand? Subway Bus Taxi 5 / 29
Approach night bus route analysis Original problem(big problem) Big data problem(small, manageable problem) Set-up mid-night buses with limited resources Floating population and moving directions of 1,252 cells Big data problem definition Modeling Floating population Night bus routes Routes optimization Optimization cases 6 / 29
Problem definition & transformation to data problems Admin. Problems Data problems 1.Where are the passengers in the mid-night? 3 billion mobile call data 2. Where do they want to go? 7 / 29
Modeling Original Simplified 605km 2 1,252 Hexagons a set of lines 8 / 29 lines with thickness
Analysis Analysis Optimization based on population 9 / 29
Analysis Finding solution through data Establish interval based on population Utilization floating population of weekdays and weekend N37 N26 10 / 29
Optimization of routes Finding answers from data - optimization <Route optimization in detail> - Route 4 : Gil-dong ~ Myung-il sta. - Route 5: Butty Hill ~ Yaksu sta. - Route 7: Express bus terminal - Route 8: Nambuterm, Konkukuniv. Route 4 95,335 105,261 (10% ) Route 5 161,483 175,233 (8.5% ) Route 790,785 95,146 (5% ) Route 8205,719 2201582 (7% ) 11 / 29
Establish 9 owl bus routes Initial routes of Seoul owl bus in 2013 12 / 29
Results Administration perspective Proof of administration decision to settle civil complaints Max. 10% of PAX increased without Wanderer due to refusal of taxi Passengers Women returning home late increasing new routes or buses 42% coverage of citizens Citizen s evaluation Student Proxy driver Daily Businessman PAX A new way to go home at mid-night 8.9% reduction of refusal rate to passengers of mid-night taxi 11.8% increase of women s mid-night activities 8.9% of refusal passengers 11.8% of midnight women s activity Source: News jelly 13 / 29
Case 2: Taxi analysis Why Taxi Matchmaking? According to 120 Seoul Dasan call center - 25.5% of the citizens complaints are on transportation! - Among them, 73.5% are related to taxis! Response of the City Provide more supplies of taxis, without additional no. of taxis - Taxi DTG (Digital Tacho-graph) -X,Y coordinate, height, date, heading, speed, status per 10 secs - Data are collected in every 150 seconds Status of taxi registered in Seoul Private Taxis 49,424 Subsidy $150M / year Corporate Taxis 22,801 14 / 29 Facing Problems It seems to be short supply of taxi during 11PM to 1AM while Taxis in Seoul are oversupplied
Approach taxi analysis Original problem(big problem) Big data problem(small, manageable problem) More taxi supply without increasing no. of taxis Decrease vacancy rate of taxis Big data problem definition Modeling Vacancy vs Demand No vacant taxis during 23-01hr Vacant rate is HIGH! Refined node/link High demand High vacancy 2 Instead of providing more taxis, what about reduction of vacancy rate? Refined link length : 150m Vacant -2 min. walking rate distance is HIGH! -Can count taxi with speed of 60km/h 3 1 Seoul taxi/pax map Share analysis results Reinforcing eco-system 15 / 29
Problem definition & transformation to data problems Admin. Problems Data problems 1.Provide more taxis without increasing no. of taxis Vacant rate = Vacant run (time or distance) Total run(time or distance) 2. Increase utilization of taxis Hired rate = Hired run(time or distance) Total run(time or distance) 16 / 29
Modeling Status of taxi 1 2 3 4 Vacant Hired Get on Get off Status of taxi Get on Get off Vacant run Hired run 1 2 Hired run 1 Vacant run 1 3 4 Hired run get off Vacant run Vacant run get off Hired run 1 1 1 1 1 1 17 / 29
Modeling Seoul metropolitan area Original Standard node/link Refined Refined node/link Average link length: 330m, Longest link length: 30km Refined link length : 150m -2 min. walking distance -Can count taxi with speed of 60km/h -(move 150m in every 10 sec.) 18 / 29
Analysis Vacancy rates show uniform patterns on a same street, but, mismatches between taxis supply and passengers demand exist on streets/roads Gangnam station B C E A D F B A C D E F A Cityhall B A D E C C B D E Hongik Univ C D F A B E A C B D F E 19 / 29
Analysis Top demand spots 연번 지역 횟수 1 홍대입구 1891 2 강남역 1654 3 가로수길 1501 4 선릉역(동) 1473 5 신림역 1316 13 12 6 건대입구 1256 7 로데오거리 1136 20 1 1km 8 압구정역 1112 9 신논현역 1109 10 선릉역(서) 1041 1km 6 17 11 사당역 979 15 8 7 21 12 동대문역 975 13 종각역 952 14 영등포역 929 14 3 19 15 이태원역 876 9 10 4 16 서울대입구 역 855 5 2 18 17 천호역 844 18 양재역 835 19 몽촌토성역 832 16 11 20 홍대입구(북) 815 21 강동역 750 기준 : 2014년 10월~11월, 금요일 23시 ~ 토요일 01시 대상 : 개인 및 법인택시(운행대수 약 34,306대) 20 / 29
Implications Example: Hongik Univ: Seogyo Hotel High demand vs. Donggyoro low demand 승차 집중 위치 승객대비 공차 많은 위치 2 3 1 Location Date Begin End period Vacant taxi Estimation Real vacancy rate 1. Hotel Seogyo 2014-12-11 23:00 23:30 42 89 47% 2. Seonghwa bd. 2014-12-11 23:50 00:20 28 37 76% 3. Hotel Yaja 2014-12-11 00:30 01:00 47 55 85% 23 / 42 21 29
Suggestion - Seoul taxi map 종각 역 동대 문역 홍대 입구 건대 입구 15 14 가로 수길 로데 오거 19 강남 역 22 / 29
What did big data find? 기준 : 4시~3시 운행 대수 Similar operation patterns between the days of the week 운행 대수 Very different operation between daytime and mid-night 23 / 29
Accelerate eco-system of big data Information of Boarding and Departing Time, Location, and Traveling Course Weather Information Node-Link Information on the Floating Population 25 / 42 24 29 Citizens utilize shared big data
Expected results Provide more chance to catch a cab by reducing empty rate of taxi We will reduce empty rate of taxi by 10% from now 27,000,000 liter / year (7M gallon / year) 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML 1ML Occupied Transfer Empty Transfer 13.9L * 44,932 ton / year 5% more chance to catch a taxi 21.2L Saving 1.5L Reduction of 1.5 liter/day CO2 * Applied IPCC formula for tco 2 conversion $40M annual cost savings - Average empty rate of taxi in Seoul = 42% (Korea Transport Institute, 2012) 26 / 42 25 29
Seoul s Big Data - Learning & Direction
What we ve learned 1.Problem definition is important! - Try to find problems by analysis? Good question makes good analysis 2. Objective of analysis? - Quest for pearl in a grain of sand? Analysis without objectives may lead wrong direction 3. Administration problems into data problems - Much better to analysis data based transformed problem *Bus route problem -> floating population & direction problem 4. How to model data? - Analysis raw data? Better to analysis with simplified but problem oriented model/data *Floating population -> population of hexagons with 500 m radius -> 605km 2 of Seoul -> 1,252 cells 5. Focused only on big data analysis? - Good insights are sometimes coming from traditional data analysis.as well as big data 6. Apply analyzed results - Provide analyzed results to proper departments. They will demonstrate results, not by you 27 / 29
Big Data Life Cycle Vision Connected.Digital.SEOUL World No. 1 Digital City Strategy Digital City Solving the city problems through Digital Digital Citizen Increase value of citizen s life by Digital Life Cycle Analyze Big Data Open & Share BIG DATA Utilize for Citizen Store on Cloud Create by IoT Area Transportation Safety Business development Health & Environment Culture & Tourism Technology Mobile Big Data GIS Communication Hub Cloud 28 / 29
What are we going to do with data? 1 2 Transport/Safety Welfare 2015 ~ -Reduction of car accidents -Optimization of local bus routes -Automatic allocation of disabled taxi -Optimization of civil service kiosks -Mobility service for the disabled 3 Small Business 4 Environment -Big data for small business develop. -Analysis for Seoul city festival -Tuberculosis analysis and more 29 / 29
Human-Centric Seoul, Happy Citizens of Seoul 사람이 중심인 서울, 시민이 행복한 서울 30 / 29
Thank you 31 / 29