Tree analysis dashboard
Results appear here after calculation.
Performance and complexity
Switch the plotted measure without rebuilding the comparison.
Selected tree structures
Pairwise comparison table
Feature usage comparison
Prediction agreement matrix
Decision-path inspector
Ensemble contribution and pruning
Statistical summary
Interpretation and recommendations
Generated tree rules
Formula used
Prediction agreement equals matching predictions divided by evaluated samples. Higher agreement means two trees behave similarly. Very high agreement may indicate redundancy.
Weighted score combines normalized performance, simplicity, speed, diversity, and stability. Each user weight controls its influence. Larger values rank the tree higher.
Diversity estimates disagreement with other ensemble members. Useful diversity can improve ensemble accuracy. Excessive diversity may signal weak trees.
How to use
Load an example or paste tree statistics. Select at least two trees. Adjust ranking and display settings.
Enter one observation for path inspection. Calculate the comparison and review recommendations. Export results in your preferred format.
Example data
| Tree | Accuracy | F1 | Depth | Nodes | Diversity | Root feature |
|---|---|---|---|---|---|---|
| Tree 1 | 0.912 | 0.911 | 6 | 43 | 0.31 | Income |
| Tree 2 | 0.887 | 0.885 | 4 | 25 | 0.46 | Age |
| Tree 3 | 0.926 | 0.926 | 8 | 69 | 0.27 | Balance |
Interpretation guide
Compare performance gains against added complexity. Inspect unique features and decision paths. Remove trees only after validation.
A representative tree resembles ensemble behavior. A diverse tree contributes different information. A redundant tree adds little new signal.
Frequently asked questions
What is an individual tree comparison plot?
It compares selected ensemble trees using performance, structure, feature usage, predictions, and contribution measures.
Why do trees in one forest differ?
Bootstrap samples and random feature selection create different splits. This diversity usually strengthens ensemble predictions.
Which tree is best?
The best tree depends on accuracy, complexity, speed, stability, and the purpose of the analysis.
Should the deepest tree be selected?
Not always. Deep trees may overfit and become difficult to explain. Validation performance should guide selection.
What does prediction agreement mean?
It measures how often two trees produce matching outcomes. High agreement can reveal redundant ensemble members.
What is a representative tree?
It produces predictions and feature patterns close to the ensemble average while remaining reasonably simple.
Can weak trees still help?
Yes. A slightly weaker tree can add useful diversity and improve the combined ensemble decision.
When should a tree be pruned?
Consider pruning when complexity increases without validation gains, or when a tree is highly redundant.
Does this calculator train real models?
It compares supplied or generated tree statistics. It also simulates structures and decision paths for learning and analysis.