FeatherCast

The voice of The Apache Software Foundation

GeoSpatial and Temporal Forecasting in Uber Marketplace Chong Sun

September 12, 2019
timothyarthur

Uber’s Marketplace is the algorithmic brain and decision engine behind our ride-sharing services. Marketplace Forecasting builds and deploys ML algorithms to handle the immense coordination, hyperlocal decision making, and learning needed to tackle the enormous scale and movement of our transportation network. In order for our decision engines to be future-aware, we need to be able to “see into the future” as accurately as possible across both space and time. In Uber, we use H3 (a hexagonal hierarchical geospatial indexing system) to partition the data geospatially. To incorporate both geospatial and temporal features into the forecasting models in real time, we need efficient and scalable techniques for data processing. In this presentation, we provide an overview of geospatial and temporal forecasting problem in Uber Marketplace Forecasting. Then, we use real time forecasting as an example to show the need for efficient data smoothing and describe a hexagon convolution based approach for processing H3 data in production. We will demonstrate how we could incorporate hexagon convolution with Apache Flink to scale the data smooth for forecasting.

Leave a Reply

Required fields are marked *.

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Blog at WordPress.com.
%d bloggers like this: