Uncertainty Estimation in Neural Networks for Time Series Prediction at Uber

Uncertainty Estimation in Neural Networks for Time Series Prediction at Uber: Reliable uncertainty estimations for forecasts are critical in many fields, including finance, manufacturing, and meteorology. At Uber, probabilistic time series forecasting is essential for accurate hardware capacity predictions, marketing spend allocations, and real-time system outage detection across millions of metrics. Classical time series models are often used in conjunction with a probabilistic formulation for uncertainty estimation. However, such models can be hard to tune, scale, and add exogenous variables to. Motivated by the recent resurgence of Long Short Term Memory networks, we propose a novel end-to-end Bayesian deep model that provides time series prediction along with uncertainty estimation at scale.