This repository contains the training and inference code for estimating whether Fujian Normal University Huaxiang Garden ChaPanda delivery orders may time out from the order time.

records.csv in the Dataset repository stores the manually maintained base records. When both timeout and non-timeout samples are present, retraining uses labeled_empirical_bayes, which is closer to a calibrated probability model. If only timeout samples are available, the artifact remains a positive-only time-risk model.

From the local conda environment:

Or download from Hugging Face:

The default output is timeout_model.json.

The output includes the estimated probability, risk level, model mode, and a calibration note. When the model mode is positive_only_time_density, treat the number as a relative risk estimate until enough non-timeout samples are collected.