Namespace: configuration
Variables
Functions
Variables
cartConfig
Const cartConfig: PartialAlgorithmConfiguration
CART algorithm default configuration - it is confusing because CART stands for Classification And Regression Trees - here we have configuration for classification tree only - gini index can not be used as split scoring function for regression trees. This term is in my opinion used incorrectly, it was invented by Leo Breiman in 1984 - check this book if you search for CART on web, majority of articles will describe tree algorithm with configuration exported here.
Remarks
possible performance problems, see onlyBinarySplits of TreeGardenConfiguration
Defined in
algorithmConfiguration/defaultCart.ts:13
Functions
getAlgorithmConfigForEachTree
getAlgorithmConfigForEachTree(dataSet, fullConfiguration): TreeGardenConfiguration[]
Function that returns configuration for each tree of random forest - it uses bagging to select set of attributes different for each tree.
Parameters
| Name | Type |
|---|---|
dataSet |
TreeGardenDataSample[] |
fullConfiguration |
TreeGardenConfiguration |
Returns
Defined in
algorithmConfiguration/randomForestConfiguration.ts:11
getSubsetOfAttributesForTreeOfRandomForest
getSubsetOfAttributesForTreeOfRandomForest(algorithmConfiguration, _dataSet): string[]
Implementation of feature bagging for random forests.
Parameters
| Name | Type |
|---|---|
algorithmConfiguration |
TreeGardenConfiguration |
_dataSet |
TreeGardenDataSample[] |
Returns
string[]
Defined in
algorithmConfiguration/randomForestConfiguration.ts:33