Delivery Time Prediction Questions

Address methods and practical challenges for predicting delivery and arrival times. Topics include spatiotemporal feature engineering, modeling traffic and preparation delays, representing agent behavior, probabilistic forecasting and uncertainty estimation using quantile regression or probabilistic models, handling censored or noisy labels and heteroscedasticity, calibration and bias correction, online updating and latency constraints for serving, evaluation metrics and calibration diagnostics, and validation strategies for production ETA systems.