© 2010 IBM Corporation
IBM Research - Ireland
© 2014 IBM Corporation
Francesco Calabrese, Yiannis Gkoufas
IBM Research - Ireland
Data driven transportation analytics
© 2010 IBM Corporation
IBM Research - Ireland
© 2014 IBM Corporation
Data-driven decision making in Transportation and City-Planning
• In the era of Big Data, authorities are provided with a rapidly increasing number of datasets
and data-streams to facilitate them in monitoring and planning on a city level
• The goal is to empower the city operators to take advantage of those data and provide
better services to the citizens
• In the last decade, a lot of city-wide sensoring systems have been installed to various cities
in the world. Our research focus is the SCATS System in the City of Dublin
© 2010 IBM Corporation
IBM Research - Ireland
© 2014 IBM Corporation
SCATS System in Dublin
• Intelligent Transportation System which involves installation of sensors on the
intersections of the City which monitor and report various KPIs related to traffic flow
(saturation, flow, green time, etc)
• The Dublin system is deployed across 550 intersections
with 3300 sensors in total and generating 150MB/hour
• Historical data are publicly available in the Dublinked
website:
• http://dublinked.com/datastore/datasets/dataset-289.php
• http://dublinked.com/datastore/datasets/dataset-305.php
• http://dublinked.com/datastore/datasets/dataset-274.php
time
© 2010 IBM Corporation
IBM Research - Ireland
© 2014 IBM Corporation
Goals of SCATS-Analytics Platform
• Leverage both open data stream and historical data obtained from Dublin's SCATS sensors
• Run analytics on the raw data and provide insights on real-time
• Present the results in a comprehensive manner, easily accessible to the City-operators
© 2010 IBM Corporation
IBM Research - Ireland
© 2014 IBM Corporation
SCATS-Analytics: Architecture Overview
• Real-time data stream is stored on a database
• The analytics run offline on fixed intervals producing the models for each detector
• Web application connects to real-time data stream, obtains the trained models and
performs online classification and detection
Historical Data
SCATS
Sensors
Trained Models
For Detectors
Analytics
WEB
Application
© 2010 IBM Corporation
IBM Research - Ireland
© 2014 IBM Corporation
SCATS-Analytics: Challenges
• The input data were unfiltered and many times contained erroneous entries
• The large amount of data volume produced made it challenging to perform analytics
• The real-time KPIs were necessary to be extracted as fast as possible in order to give the
City-operator prompt input for decision-making
• Finally, the results of all those advanced analytics should be presented to the domain
expert in a user-friendly and comprehensive way
© 2010 IBM Corporation
IBM Research - Ireland
© 2014 IBM Corporation
SCATS-Analytics: State Classification Scheme
Free flow F
Congestion C
Transient states U
(due to “network-effects”)
Optimal
service rate
Characterization of network traffic processes under adaptive traffic control systems
A. Pascale, T.L. Hoang, R. Nair
21st International Symposium on Transporation and Traffic Theory, 2015
© 2010 IBM Corporation
IBM Research - Ireland
© 2014 IBM Corporation8
SCATS-Analytics: City-wide overview
© 2010 IBM Corporation
IBM Research - Ireland
© 2014 IBM Corporation9
SCATS-Analytics: Historical data inspection
© 2010 IBM Corporation
IBM Research - Ireland
© 2014 IBM Corporation
SCATS-Analytics: Lessons Learned
• Available Open Data are useful to be incorporated in the process of decision-making, but
domain knowledge is required to fully leverage on them.
• Investing on improving the quality and the availability of data sources can only be beneficial
for the citizens
• When designing a platform, it's necessary to keep the end-user involved in the process of
the development, looking for feedback and in the end creating a high-level user experience
© 2010 IBM Corporation
IBM Research - Ireland
© 2014 IBM Corporation
THANK YOU
Questions?

Data Driven Tranportation Analytics

  • 1.
    © 2010 IBMCorporation IBM Research - Ireland © 2014 IBM Corporation Francesco Calabrese, Yiannis Gkoufas IBM Research - Ireland Data driven transportation analytics
  • 2.
    © 2010 IBMCorporation IBM Research - Ireland © 2014 IBM Corporation Data-driven decision making in Transportation and City-Planning • In the era of Big Data, authorities are provided with a rapidly increasing number of datasets and data-streams to facilitate them in monitoring and planning on a city level • The goal is to empower the city operators to take advantage of those data and provide better services to the citizens • In the last decade, a lot of city-wide sensoring systems have been installed to various cities in the world. Our research focus is the SCATS System in the City of Dublin
  • 3.
    © 2010 IBMCorporation IBM Research - Ireland © 2014 IBM Corporation SCATS System in Dublin • Intelligent Transportation System which involves installation of sensors on the intersections of the City which monitor and report various KPIs related to traffic flow (saturation, flow, green time, etc) • The Dublin system is deployed across 550 intersections with 3300 sensors in total and generating 150MB/hour • Historical data are publicly available in the Dublinked website: • http://dublinked.com/datastore/datasets/dataset-289.php • http://dublinked.com/datastore/datasets/dataset-305.php • http://dublinked.com/datastore/datasets/dataset-274.php time
  • 4.
    © 2010 IBMCorporation IBM Research - Ireland © 2014 IBM Corporation Goals of SCATS-Analytics Platform • Leverage both open data stream and historical data obtained from Dublin's SCATS sensors • Run analytics on the raw data and provide insights on real-time • Present the results in a comprehensive manner, easily accessible to the City-operators
  • 5.
    © 2010 IBMCorporation IBM Research - Ireland © 2014 IBM Corporation SCATS-Analytics: Architecture Overview • Real-time data stream is stored on a database • The analytics run offline on fixed intervals producing the models for each detector • Web application connects to real-time data stream, obtains the trained models and performs online classification and detection Historical Data SCATS Sensors Trained Models For Detectors Analytics WEB Application
  • 6.
    © 2010 IBMCorporation IBM Research - Ireland © 2014 IBM Corporation SCATS-Analytics: Challenges • The input data were unfiltered and many times contained erroneous entries • The large amount of data volume produced made it challenging to perform analytics • The real-time KPIs were necessary to be extracted as fast as possible in order to give the City-operator prompt input for decision-making • Finally, the results of all those advanced analytics should be presented to the domain expert in a user-friendly and comprehensive way
  • 7.
    © 2010 IBMCorporation IBM Research - Ireland © 2014 IBM Corporation SCATS-Analytics: State Classification Scheme Free flow F Congestion C Transient states U (due to “network-effects”) Optimal service rate Characterization of network traffic processes under adaptive traffic control systems A. Pascale, T.L. Hoang, R. Nair 21st International Symposium on Transporation and Traffic Theory, 2015
  • 8.
    © 2010 IBMCorporation IBM Research - Ireland © 2014 IBM Corporation8 SCATS-Analytics: City-wide overview
  • 9.
    © 2010 IBMCorporation IBM Research - Ireland © 2014 IBM Corporation9 SCATS-Analytics: Historical data inspection
  • 10.
    © 2010 IBMCorporation IBM Research - Ireland © 2014 IBM Corporation SCATS-Analytics: Lessons Learned • Available Open Data are useful to be incorporated in the process of decision-making, but domain knowledge is required to fully leverage on them. • Investing on improving the quality and the availability of data sources can only be beneficial for the citizens • When designing a platform, it's necessary to keep the end-user involved in the process of the development, looking for feedback and in the end creating a high-level user experience
  • 11.
    © 2010 IBMCorporation IBM Research - Ireland © 2014 IBM Corporation THANK YOU Questions?