cdr data analysis for epidemic control (itu) 20160126 · cdr data analysis for epidemic control itu...
TRANSCRIPT
CDR Data Analysisfor Epidemic Control
ITU Technical Team for CDAECProf. R. Shibasaki,Dr. H.Kanasugi,
Dr.A.Watayangkurn andDr.W. Ohira
2015.08.26SHIBASAKI & SEKIMOTO LAB. CSIS/IIS/EDITORIA. UTOKYO 3
Estimated People Movements (2015.06.07) N=640,504
Entire Sierra Leone
Freetown Bo
4
People Flow (7, June 2015) Sierra Leone
Freetown(CapitalCity)
Kenema(Hospital)
Possibleepicenter?
MapofSierraLeone
Contribution of CDR Analysis• Real-time Information on People Flow will help
– Ebola outbreak analysis and control– Evacuation Guidance from Floods– Road Planning and Management etc.– Transboundary movement could be estimated by using IMSI, if
randomized with the same seed number.• For MNOʼs
– Base station optimization etc.
6
PrivacyisProtectedwithRandomization
7
IMSI,IMEIetc
Identificationinformation
connectedtoaspecificuser
RandomizedID
Randomized,uniquenumber
NOT connectedtoaspecificuser
Randomization
RegulatoryAuthorities(Analyst@ITU)
Impossible torestoreIMSIandIMEIfromthe
RandomizedID.
ByMNO(MobileNetworkOperator)
AggregatedAnalysisResults
CDR(CallDetailRecord)data
8
data
data
Dataontimeandlocation(basestation)ofvoice/messaging/datacommunicationofeachhand-setforbillingpurposes.Dataisrecordedonlywhencalling,messaginganddatacommunication.
AnImportantFinding
• Shiftingfromruralareastocitiesandtownsfortheoutbreak.– TogetherwithPeopleFlow.
HowtoTrackPeopleFlow?
• QuestionnaireorInterview?
• Countingvehiclesatgatesorsurveillancepoints?
• Cellularphonedata(CDRanalysis)!
HowtoIdentifytheLocationofSubscribers?
Source:http://www.itisholdings.com/article.asp?id=296http://5.freshminutes.it/2008/05/20/geolocalisation-mobile-computing/
Cell-IDLocation:<coarse>100m.– 4km.
11
GPSLocation:<good>10m.– 200m.
Needtoembedlocationdatatransmissionsoftware
Allhand-setscanbecoveredwithoutanyadditional softwareforhand-sets.
Allhand-sets
Movementofahandset(CELLTOWERMigration)
MappingMovementofaHandsetfromCDRdata
Staypoints
Estimatingrouteandstaypointsofasubscriberthroughinterpolation.
è RepeatingforallN
subscribers
EstimatingMovingRouteandStayPointsofaSubscriberfromtheMovementoftheHandset
Move
Stay
Time
Space0:00am
6:00am
12:00pm
Aggregation EstimatedMovingRouteandStayPointsofaSubscriber
EstimatingHourlyPopulationDensitybyAggregatingNumberofPeopleinaGrid-Cell.
Needtoestimatescalingfactor to covertsubscribernumber tototalpopulationAt8:00AM
CallingDemandDensityataspecifictimeslotbyusingpopulationdensityandcallingbehaviormodel
Needtoestimatecallingdemandoftotalpopulationbasedon“callingbehaviormodel”.
Optimizingcelltowerlocation
SHIBASAKI & SEKIMOTO LAB. CSIS/IIS/EDITORIA. UTOKYO 17
People Flow (7, June 2015) Freetown
2015.08.26SHIBASAKI & SEKIMOTO LAB. CSIS/IIS/EDITORIA. UTOKYO 18
Hourly Population Density (7, June 2015) Sierra Leone
2015.08.26SHIBASAKI & SEKIMOTO LAB. CSIS/IIS/EDITORIA. UTOKYO 19
Hourly Population Density (7, June 2015) Freetown
SHIBASAKI & SEKIMOTO LAB. CSIS/IIS/EDITORIA. UTOKYO
20
Hourly Population Density (7, June 2015) Freetown
Further Analysis of People Flow; Freetown Bo
Suggestionsforfutureworks
• 2nd Pilotproject– Trackingtransboundary(international)movementofpeopleamongSierraLeone,GuineaandLiberia.
– ConnectingdatawithRandomizedIMEI.
• DevelopmentofReal-timeMonitoringandAnalysisSystemforSierraLeone,GuineaandLiberia– MNO’s:RandomizingCDR– RA’s:IntegratingRandomizedCDRandConductingAnalysistoMapReal-TimePeopleFlow
– ITU:TechnicalSupportandCapacityBuilding21
OriginalCDRdataprovision
Randomized/Anonymized CDRdata
Aggregating forPeopleflowstatistics
Additional analysisforEbolaControl
Reporting
ByMNO
ByMNO
ByMNOwithtechnicalassistanceofUT
ByUTunderthesupervisionofITU
ByUTunderthesupervisionofITU
ConductedinthepremiseofMNO
underthesupervisionofNationalRegulatoryCommissionandITU
Peopleflowstatistics
22
NationalRegulatoryCommission
Supervision
Peopleflowstatisticsdata
Appendix;DataManagementSchemeInthepremiseofMNO
全体の流れ
• 課題–エボラは⼈の移動にともなって動く–⼈の移動に関するデータが無い。
• ソリューション– CDR分析を⾏う。
• 効果、リスク(対応⽅法)–⽇常的に得られるデータで、安価にリアルタ
イムに⼈の移動データが得られる。–携帯電話事業者内でRandomizeし、統合解析
はRAが⾏うことで、プライバシーリスクは避けられる。