How Predict works — end-to-end
This is the daily loop in the UI.
Step by step
A. Get in
Sign in → confirm AI Project → open Predict (AIAAS). Check Dashboard so you know if jobs already exist.
B. Prepare data
- Prefer a clean CSV with one date/time column and one numeric target.
- Data Sources lists prior uploads; new files are added when you run + New prediction (or via API
analyze_csv/predict).
C. Forecast
- Prediction Jobs → + New prediction.
- Pick model, prediction type, periods, date column, target column, CSV.
- Wait for SUCCESS. Note MSE/RMSE and model path.
D. Review
Predictions for charts/tables/export. Dashboard for portfolio counts. Audit Logs for “who did what”.
E. Clean up
Trash icon on a job removes that job (and temp artifacts when unused). Failed jobs stay visible until you delete them — useful for debugging.
What MSE / RMSE mean
| Word | Meaning for you |
|---|---|
| MSE | Mean squared error on training (or hold slice) |
| RMSE | Square root of MSE — same units as your target, easier to read |
| Accuracy % | Not computed by Predict today |
| Prophet / ARIMA / XGBoost | Algorithms (“High accuracy” in some UIs = Prophet) |
If status is FAILURE, open the job error (SSL/DB, bad columns, empty training range). See Troubleshooting.