R-Boosting Classification

Properties that can be configured for the R-Boosting Classification algorithm.

Overview:
The Boosting algorithm is a popular ensemble method that can be applied for classification. The Adaboost.M1 and Adaboost-SAMME algorithms are supported in the component. The ensemble method is designed to improve the accuracy and robustness of weak classifiers on business datasets.

The R packages that implement the algorithm are adabag and rpart.

Note

In this component, the decision tree method is selected as the classification algorithm.

Note

When the column names contain the hyphen symbol (-), use the Data Type component to re-define the column name.

R-Boosting Classification Properties
Table 1: Algorithm Properties
Property Description
Maximum Depth Enter the maximum node level in the final tree with the root node counted as level 0. This parameter can be set between 1 and 20 inclusive.
Minimum Split Enter the minimum number of observations required for splitting a node. The default value is 0. The parameter can be set between 0 and 500 inclusive.
Complexity Parameter Enter the complexity parameter, which saves computing time by preventing any split that does not improve the fit. The value for the parameter must be between [-1, 1), which is equal to or more than -1 and less than 1.
Number of Iterations Number of iterations for which boosting is running. This parameter can be set between 5 and 500 inclusive.
Sample Weights If TRUE, a bootstrap sample of the training set is drawn by using the weights for each observation on that iteration. If FALSE, every observation is used with its weights.
Weight Updating Coefficient Three ways to calculate the weight updating coefficient, which is α in AdaBoost.M1 algorithm are as follows: A) ‘Breiman’: α=1/[2 ln⁡((1-err)/err)], and B) ‘Freund’: α=ln⁡((1-err)/err), and⁡ C) ‘Zhu’: α=ln⁡((1-err)/err)⁡+ln⁡(N_classes-1).
Features Select the input columns with which you want to perform the analysis.
Target Columns Select the target column on which you want to perform the analysis.