Table 1.

Summary of the pipeline methods and steps for online anomaly detection using surveillance data.

Step No.	Major Activities
1	Train candidate anomaly detectors per health region using train set
2	Based on local epidemic demands, select best anomaly detector, $M_{best}$
3	Tune model parameters after using for 6–12 months to evaluate performance
4	As new data arrives, use the best detector, $M_{best}$ to detect and interpret anomaly
5	New data for evaluation and model re-training
6	Update the models with the new data and repeat from Step 1