¶ Description
Create a ROC curve after applying a model.
¶ Parameters
¶ Parameters tab
Parameters:
- (Optional) Target Name
- (Optional) Variable Suffix
- Label Language
- Select TARGET variable
- Select PROBABILITY Variable(s)
- Maximum amount of rows to plot
- Select Line Color
- *Select Area Color for Single Plots
- Area Color for Comparison Plots
- Number of Charts per Rows
- PNG Plot size (in Pixel)
- Individual Plots
- Comparison Plots
- "All In One" plots
- Plot Accuracy
- Precision Digits for AUC
- Alpha transparency (area)
- All Data
- Variable To Select
¶ Description tab
Parameters:
- Script name
- Short description
- Revision
- Decription
¶ Configuration tab
See dedicated page for more information.
¶ About
The R_ROC action button generates a Receiver Operating Characteristic (ROC) Curve based on predicted probabilities and binary classification targets. This tool evaluates model performance by plotting the True Positive Rate (TPR) against the False Positive Rate (FPR) for different probability thresholds. The Area Under the Curve (AUC) metric quantifies the model's classification ability.
¶ Input
-
Format:
.csvtable -
Required Columns:
Target: Binary values (0 or 1), representing the true classProba: Predicted probability values from the model (between 0 and 1)
-
Minimum data requirement: At least 100 rows for stable and meaningful ROC curve generation. While it technically works with fewer rows, results may be misleading (e.g., AUC = 1 with diagonal curve).
-
Example Input Table:
| Target | Proba |
|---|---|
| 1 | 0.95 |
| 0 | 0.10 |
| 1 | 0.88 |
| ... | ... |
¶ Parameters
| Parameter | Description | Example / Value |
|---|---|---|
ModelName |
Optional label for target model | Model_A (optional) |
idSuffix |
Optional suffix for variable naming | _v1 (optional) |
idLanguage |
Interface language for plot | English |
Target |
Target column (binary) | Target |
Proba |
Probability column | Proba |
idMaxRows |
Max number of rows to process | 50000 |
myColor1 |
ROC line color | #00007f |
ColorArea |
Area fill color for individual plot | #5500ff |
ColorAreaAll |
Fill color for comparison plots | #2bc7ff |
idCol |
Number of plots per row | 3 |
idSize |
Plot image size in pixels | 800 |
idOnePlot |
Enable individual plots | true |
idAllPlot |
Enable comparison plot | false |
idComboPlot |
Enable combo plot (all-in-one) | false |
iRound |
Decimal precision for AUC | 5 |
iAlpha |
Area transparency | 0.1 |
idOutput |
Output format | chart |
idxTitle |
Chart title (for HTML report) | Weight ROC Plot |
¶ Output
-
Image file: ROC curve (
ROC_Proba.png) -
Content:
- ROC curve (TPR vs FPR)
- AUC score displayed on the plot
-
Example ROC Output:
¶ Use Cases
- Evaluate binary classifier models such as logistic regression, random forest, XGBoost, etc.
- Compare classifier performance by AUC metric
- Visualize model discrimination capability at different thresholds
¶ Common Pitfalls & Troubleshooting
¶ ❌ AUC = 1 with diagonal ROC curve
- Cause: Too few rows or repeated identical values in
Proba - Fix: Use at least 100+ rows with varied probability values
¶ ❌ AUC = 0.5 or meaningless results
- Cause: Target column is constant or incorrectly formatted
- Fix: Ensure
Targetcontains both 0 and 1 classes
¶ ❌ "Too many probability levels" error
- Cause: Excessive unique values in
Probacolumn - Fix: Limit
Probato fewer than 1200 unique values. Round to 2–4 decimals if needed.
¶ How to Run
- Prepare input CSV with
TargetandProbacolumns. - Configure parameters (optional: set language, color, plot style).
- Run pipeline and retrieve output
.pngfile from the log or records tab.
¶ Sample Pipeline
