Crowd-source data to activity models: Human mobility prediction for real-time ride-sharing

Crowd-source data to activity models: Human mobility prediction for real-time ride-sharing
PI: R. Sengupta, UC Berkeley 
$92,405

Abstract: We focus on the crowd‐sourcing of location and text data from sources like Google Calendar, text messages, Facebook and tweets. We propose to stream this data in quasi‐real time for three months from each of 200 people to the cloud and use it to predict the future destinations and activities of the person streaming the data. Our emphasis on text and activity is to enable all of the data collected for demand modelers by survey to be crowd‐sourced. Since the data streams in real time, it enables models that predict the future locations and activities of a traveler, i.e., a Human Mobility Model. These models bring reliability and trust to real‐time ride‐sharing – a known and important need. We aggregate the individual mobility models to estimate the number of people expected to be at a particular place at a future time and its standard deviation. Thus such crowd‐sourced data can help transportation system managers estimate demand in real‐time if it becomes available for a few thousand people.

Quarterly Progress Reports

Quarterly progress report for the 1st quarter of FY 2014-15.

Quarterly progress report for the 2nd quarter of FY 2015-16.