01
This is a purely static educational site. “AI prediction” names a model-evaluation topic; it does not mean a prediction API has been deployed.
02
The simulator uses browser-side JavaScript and random sampling. It does not call probability arithmetic a trained model or package random numbers as real predictions.
03
Any future model would first need lawful data sources, field definitions, timestamps, rule versions, missing-data handling and features available at deployment.
04
Training, validation and test sets should follow time order. Preprocessing must be fitted on the training period, and the final test must stay unseen until the design is locked.
05
Compare the model with majority-class, rules-based or simple-frequency baselines, and report sample size, intervals, calibration and all failures.