Real-Time Spatial Temporal Forecasting @ Lyft

110 · Lyft · May 5, 2025, 5:33 p.m.
Summary
This blog post discusses real-time spatial temporal forecasting at Lyft, detailing the models, challenges, and implementation strategies used to predict market conditions for ridesharing services. It compares classical time-series models and deep neural networks (DNNs), illuminating the trade-offs in accuracy, latency, and computational costs based on the specific characteristics of ride demand and supply signals.