METRANS researchers Dr. Cyrus Shahabi and Dr. Ugur Demiryurek, both of the Integrated Media Systems Center at USC’s Viterbi School of Engineering, recently published their research findings. This project is one of the METRANS projects funded in full by Caltrans. The research addressed the problem of real-time traffic prediction by developing algorithms to deal with missing values and missing sensors for large road networks. The final report is titled “Analysis and Prediction of Spatiotemporal Impact of Traffic Incidents for Better Mobility and Safety in Transportation Systems.”

Installed traffic sensors enable researchers and planners to predict traffic and to improve traffic management, regulations, and urban planning with the collected traffic parameters such as occupancy, volume, and speed. More than 15,000 loop detectors are installed on the Los Angeles County roadway network, both highways and main corridors. However, not every street is equipped with such monitoring sensors. Sensors are also subject to occasional failure. These conditions mean the dataset collected from sensors is incomplete. This research aims to solve this data problem by studying datasets from the past five years for many large cities.

Shahabi and Demiryurek proposed a Latent Space Modeling for Road Networks (LSM-RN) to achieve more accurate and scalable traffic prediction, which utilizes both topology similarity and temporal correlations. Later, both global learning and incremental learning algorithms were designed to balance the accuracy and efficiency of real-time forecasting. Experiments were conducted to prove that LSM-RN could obtain higher accuracy than existing time series methods. Additionally, this model is also capable of handling larger networks. Results show that it only takes 4 seconds to make a prediction for a network with 19,986 edges.

A two-page summary of the report can be found here.

The full report can be found here.

P.I.:

Cyrus Shahabi 

Professor and Director of the Information Laboratory (InfoLAB), Department of Computer Science; USC Viterbi School of Engineering, University of Southern California

3737 Watts Way, Charles Lee Powell Hall (PHE) 306A

Los Angeles, CA 90089-0781

[email protected]

Co-P.I.:

Ugur Demiryurek 

Research Scientist, Integrated Media Systems Center

USC Viterbi School of Engineering, University of Southern California

3737 Watts Way, Charles Lee Powell Hall (PHE) 335

Los Angeles, CA 90089-1211

[email protected]

Author

Kaitlyn (Kenan) Zhang

Kaitlyn is a first-year Master of Public Policy (MPP) student attending the Sol Price School of Public Policy at the University of Southern California. Her interests are in transportation policy, urban development, and economics. She is also working towards a certificate in Transportation Systems from the Viterbi School of Engineering.