Results
Selected node
- Select a node in the diagram.
Feature importance
Performance metrics
Decision rules
No rules generated.
Prediction path
Node and branch table
| ID | Depth | Type | Rule | Samples | Impurity | Gain | Prediction |
|---|---|---|---|---|---|---|---|
| No model trained. | |||||||
Dataset and model settings
Formula used
Gini impurity: Gini = 1 − Σp². Lower values indicate purer classification nodes.
Entropy: H = −Σp log₂(p). Information gain subtracts weighted child entropy from parent entropy.
Regression error: MSE = Σ(y − ȳ)² / n. Splits maximize weighted error reduction.
Feature importance: each feature receives its accumulated impurity reduction, weighted by node sample share.
How to use
- Paste data, upload CSV, or edit the manual table.
- Select the target column and task type.
- Choose depth, leaf, split, pruning, and display settings.
- Build the tree and inspect nodes, rules, metrics, and paths.
- Enter new values to highlight a prediction route.
- Export the diagram, node table, rules, or complete model.
Example data
| Outlook | Temperature | Humidity | Windy | Play |
|---|---|---|---|---|
| Sunny | 85 | 85 | False | No |
| Overcast | 83 | 78 | False | Yes |
| Rain | 70 | 96 | False | Yes |
| Sunny | 69 | 70 | False | Yes |
Frequently asked questions
Each node shows its rule, samples, impurity, gain, and prediction.
Yes. It evaluates categorical equality splits and numeric thresholds.
Choose row removal, imputation, or a fixed branch direction.
Excessive depth and tiny leaves often memorize training observations.
Pruning removes weak branches to improve generalisation and clarity.
Yes. Numeric targets can use MSE, MAE, or variance reduction.
Importance accumulates weighted impurity reductions from every selected split.
Yes. Prediction inputs highlight every branch taken through the tree.
CSV input is supported, with PNG, SVG, PDF, CSV, JSON, and text exports.