Enabling Demand Modelling from Privately Held Mobility Data
PI: Alexei Pozdnukhov, UC Berkeley
Abstract: Novel mobility paradigms such as car- and ride-sharing, on-demand transportation and proliferation of non-motorized modes change the transportation landscape quicker than traditional data sources, such as travel surveys, are able to reflect. At the same time, valuable mobility data are locked in private repositories of telecoms and service providers and cannot be easily shared whether for profit or objectives of public good. There exists a combination of customer privacy and security issues, the lack of business models for data-centric private-public partnerships, immaturity of data marketplaces and significant technical bottlenecks that are in the way. Paradoxically, this creates a data drought in the age of the big data deluge. Public agencies charged with a mandate to manage critical transportation infrastructures lack instruments to investigate and react to rapidly evolving demand patterns.
In this project, we will research how travel demand modeling can be enhanced and demand forecasting latency reduced within the current operation practices. This project will develop and showcase a novel algorithmic framework to calibrate discrete choice models using data held by the private sector in an efficient and privacy-preserving way that does not require disclosures of sensitive personal data.