¶ Description
Compute an univariate spline regression.
¶ Parameters
¶ Parameters tab
Parameters:
- Input table partitioning (1st pin)
- Dependent variable(x)
- Target variable(y)
- Split knots definition
- Mod: number of quantile splines
- Mod: type of quantiles
- Mod: manual knots (separate by comma)
- Mod: quadratic degree
- Mod: missing values
- Lbl: thousands separator
- Lbl: rmse/r2 precision
- Saved model criteria
- Out: save model directory
- Out: model file: include time stamp in name
- OUT: save model(s)
¶ Description tab
See dedicated page for more information.
¶ Configuration tab
See dedicated page for more information.
¶ About
The R_Spline action performs univariate spline regression using a numeric predictor variable (x) and a target variable (y). It fits a smooth, non-linear curve to the data by applying piecewise polynomials joined at "knots." The result includes a spline regression plot, performance metrics (RMSE, R²), and a serialized model file.
¶ Input
- Input Table (1st Pin):
- Must include:
- One numeric predictor column (e.g.,
x) - One numeric target column (e.g.,
y)
- One numeric predictor column (e.g.,
- Minimum: 3 rows for basic spline fitting
- Format: CSV or ETL table
- Must include:
Example Input Table:
¶ Output
-
📈 Plot Image:
records/Spline_y.png– spline regression curve -
💾 Model File:
records/Spline_y_by_x_A.rModel– serialized R model for future use -
Log:
Contains:- Spline formula
- RMSE, R², Adjusted R²
- Spline coefficients
- Knots
¶ How to Run
-
Upload the File
Load your dataset using thereadCSVaction and connect it toR_Spline.
Placeholder: screenshot showing readCSV → R_Spline → RunToFinishLine -
Configure Parameters
Set the predictor, target, and spline settings. Enable model saving if needed.
-
Run the Pipeline
Execute the pipeline. -
Review Outputs:
-
Log Tab:
-
Records Tab:
Spline_y.png: Final spline plot
Placeholder: spline regression chartSpline_y_by_x_A.rModel: Serialized model file
-
¶ Example Output
Log Summary:
RMSE = 0
R² = 100
Adjusted R² = 100
Knots: [1, 5, 10]
Formula:
y ~ s(x)
¶ Use Cases
-
Scientific Data Modeling
Fit smooth experimental curves in physics, chemistry, or engineering datasets. -
1D Time-Series Smoothing
Model non-linear temporal trends (e.g., temperature over time). -
Feature Engineering
Transform non-linear signals into smooth trend-based features for use in other models.
¶ Common Errors & Fixes
| Error | Cause | Solution |
|---|---|---|
ggsave(): Cannot find directory |
Output directory doesn't exist | Use simple folder like results under records/ |
Failed to save Model File |
Invalid idxSaveDir path |
Ensure records/results/ exists |
Not enough data to run a model |
Too few rows in dataset | Minimum 3 rows, ideally ≥ 10 |
¶ Tips & Notes
-
Manual Knots:
To manually define knots, setidMethodtomanualand define them inidmKnots. -
Avoid Folder Errors:
Don’t use spaces or special characters inidxSaveDir. -
Perfect Scores Warning:
IfRMSE = 0andR² = 100, your spline may be overfitting—especially with small datasets.
