Grad Student Seminar: Yunxiao Liu, Uber
Spatio-Temporal Forecasting at Uber
In this talk I am going to give a high-level introduction to the data science projects we have been working on at Marketplace Forecasting at Uber. Marketplace Forecasting at Uber stands at the upstream of the marketplace optimization flow and generates time series and machine learning models for quantities like supply, demand and marketplace balance for Uber. These forecasts, at varying levels of spatial/temporal granularity and with different forecast horizons, provide a forward view of marketplace and empower downstream optimization algorithms such as pricing and incentive allocation in a real-time manner. These forecasts aim to model the world and proactively capture calendar patterns and shocks (holidays, events, weather and more) using a combination of classical time series and statistical and machine learning approaches in a scalable way.