The function produces a model and a JSON representation of the decision tree. The model output is as follows:
| Column Name | Data Type | Description |
|---|---|---|
| task_index | SMALLINT | The AMP that produces the decision tree. |
| tree_num | SMALLINT | The identified decision tree within an AMP. |
| regression_tree or classification_tree | VARCHAR | The JSON representation of decision tree. If the JSON string is larger than 32000 bytes, then it is truncated and displayed in several rows associated with a tree_order id. |
The JSON representation of the decision tree has the following elements:
| JSON Type | Description |
|---|---|
| id_ | The unique identifier of the node. |
| sum_ | [Regression trees] The sum of response variable values in the node. |
| sumSq_ | [Regression trees] The sum of squared values of the response variable in the node. |
| responseCounts_ | [Classification trees] The number of observations in each class of the node. |
| size_ | The total number of observations in the node. |
| maxDepth_ | The maximum possible depth of the tree starting from the current node. For the root node, the value is max_depth. For leaf nodes, the value is 0. For other nodes, the value is the maximum possible depth of the tree, starting from that node. |
| split_ | The start of JSON item that describes a split in the node. |
| score_ | The GINI score of the node. |
| attr_ | The attribute on which the algorithm splits the node. |
| type_ | Type of tree and split. Possible values:
|
| leftNodeSize_ | The number of observations assigned to the left node of the split. |
| rightNodeSize_ | The number of observations assigned to the right node of the split. |
| leftChild_ | The start of the JSON item that describes the left child of the node. |
| rightChild_ | The start of the JSON item that describes the right child of the node. |
| nodeType_ | The node type. Possible values:
|