Skip to main content
AI, Data / ML

DeepETA: How Uber Predicts Arrival Times Using Deep Learning

10 February 2022 / Global
Featured image for DeepETA: How Uber Predicts Arrival Times Using Deep Learning
Figure 1: Hybrid approach of ETA post-processing using ML models
Figure 2: An overview of the DeepETA model pipeline
Figure 3: Attention matrix in self-attention
Figure 4: An illustration of pairwise dot products of input features: colors correspond to attention heads and color shares correspond to random generated attention weights
Figure 5: An illustration of multi-resolution location grids with multiple feature hashing using independent hash functions h1 and h2
Figure 6: Illustration of the DeepETA model structure
Figure 7: Asymmetric Huber Loss
Figure 8: Model Re-training and deployment pipeline
Figure 9: High-level system diagram of online serving
Xinyu Hu

Xinyu Hu

Xinyu Hu is a Senior Research Scientist at Uber AI focused on large-scale machine learning applications in spatial-temporal problems and causal inference. Xinyu holds a Ph.D. in Biostatistics from Columbia University.

Olcay Cirit

Olcay Cirit

Olcay Cirit is a Staff Research Scientist at Uber AI focused on ML systems and large-scale deep learning problems. Prior to Uber AI, he worked on ad targeting at Google.

Tanmay Binaykiya

Tanmay Binaykiya

Tanmay Binaykiya is a Senior Software Engineer on the Maps ETA Prediction team, and focuses on projects at the intersection of ML and systems.

Ramit Hora

Ramit Hora

Ramit Hora is the Technical Program Management Lead for Uber AI & Maps where he works to drive innovation across Uber’s technical systems and product experiences.

Posted by Xinyu Hu, Olcay Cirit, Tanmay Binaykiya, Ramit Hora

Category: