Big Data In, Travel Insights Out
Big Data–powered machine learning reveals where people go and why
January 27, 2021: The SACOG travel-modeling team has a powerful new source of data that enables even more fine-tuned analysis of travel trends in the Sacramento region.
In a Big Data pilot project partnership with Caltrans and the California Air Resources Board, the team has access to the software-as-a-service platform Replica for 2021. Replica uses machine learning to turn many streams of inputs — raw data from GPS to cell phone apps to credit card purchases, and many more — into valuable insights useful for transportation modelers, planners, and policy-makers.
These insights are displayed in two different forms on the Replica platform: Nationwide Trends and Local Places. The Nationwide Trends is a weekly snapshot of activities across the United States such as numbers of daily trips, vehicle miles traveled, trip start times, trips to work and school, as well as spending across sectors such as gas stations, grocery stores, and restaurants and bars. Users can view the trends at the state, county, and census tract levels.
The tool uses census data and other sources about who lives in a given area, regional housing and employment availability, and then applies modeling and optimization algorithms to generate a representative population that is statistically equivalent to the census population, and match the population to personas. These representative people and households are assigned housing units and locations of workplaces and schools. Personas extract behavioral patterns from de-identified mobile location data collected from mobile devices of real people. In this way, the source data is aggregated and de-personalized so the resulting insights are anonymized.
The first step for planners using Replica is to validate the tools synthetic data by comparing it to on-the-ground observed data. SACOG’s modeling team validated it against Caltrans traffic data, population data, and data from local agencies and found it to be good for a variety of applications. Replica also tags its own insights with measures of how confident it is about the data.
Now that SACOG is confident about Replica’s data, it can be used by member and partner agencies to gain insights about their own projects. Users can create case studies in the Local Places feature, which shows information in quarters from fall 2018 to fall 2020. For example, using filters and layers, planners can do road link analysis and transit line analysis, showing information such as trip origin, destination length of trip, trip mode, ethnic breakdown of travelers, household income of transit line users, and even the reason for the trip.
The ability to slice and dice the data in so many ways makes Replica a powerful tool for understanding the impact of transportation projects and land use, and will be helpful for supporting planning grant applications. As SACOG’s Data and Analysis Manager Bruce Griesenbeck put it, “Big Data holds promise to help transportation planners to better understand how travelers are adapting to all the changing world we live in.” He said the modeling team was learning where Big Data was proving useful and just as important, where it may not be.
For partners and members who want to explore the Big Data insights of Replica, the modeling team is eager to help you get started and to help you understand if selected data is useful for your projects. To learn more and to get access to set up your own account, contact data analyst Shengyi Gao.