| English | 中文 |
A single tree partitions the predictor space into local regions. Its rules can be inspected, but small data changes can alter the partition. Bound depth, node/leaf support and pruning complexity against noise, independent sample support and deployment coverage. Select these controls using development evidence. A split in a fitted tree is a predictive association, not a causal mechanism.
Parameter reasoning and cited evidence: model comparison.
The R and Python implementations expose different controls and need separate operational contracts; matching numeric values does not establish equivalent trees. Configuration, code calls and current checks belong to the matching skill.
Use the shared evaluation design and submit a model plan before generating experiment code.