1. Load stock data
No CSV loaded
Required CSV format
Each row represents one day. Use numeric value columns.. Each selected input receives one weight that applies to every day.
2. Choose inputs and outputs
An input may influence the stock (coal, steel, dollar rate). An output is what you want to predict (closing, high or low price).
Selected parameters and one-time weights
Enter each parameter’s weight once. It applies to all days. In optimized mode it initializes training; in fixed mode it remains unchanged.
No parameters selected. Choose a column above and click Add parameter.
Outputs to predict
The program trains a separate model for each output.
No outputs selected.
3. Results
Optimized parameter weights
Training loss
Actual vs predicted
What do these results mean?
- Degree
- Curve complexity. Degree 1 is a straight line; higher degrees follow bends but may overfit.
- Terms
- Mathematical combinations used, including interactions such as Coal Price × Dollar Rate.
- RMSE
- Typical prediction error, with larger mistakes penalized more. Lower is better; it uses the output price unit.
- MAE
- Average absolute prediction error. Lower is better and it uses the output price unit.
- R²
- Variation explained. Near 1 is strong, near 0 is weak, and below 0 is worse than using the average.
- Validation
- Data used to choose the best curve without training on it.
- Test
- Untouched data used only for the final performance check.
- Weight share
- Relative emphasis on each input. It shows association, not proof of cause.
4. Make a prediction
Important
This educational tool finds mathematical relationships in historical data. Polynomial fits can overfit and stock markets change. Do not treat a fitted value as investment advice or a guaranteed future price.