ImbTreeEntropyInter: Fit an Interactive Decision Trees

In KrzyGajow/ImbTreeEntropy: Software to build Decision Trees for imbalanced data

Usage

ImbTreeEntropyInter( Y_name, X_names, data, depth = 5, min_obs = 5, type = "Shannon", entropy_par = 1, 
                     cp = 0, n_cores = 1, weights = NULL, cost = NULL, class_th = "equal", 
                     overfit ="leafcut", cf = 0.25, amb_prob = 1, top_split = 2, var_lev = T, 
                     amb_class = NULL, amb_class_freq = NULL, tree_path = getwd() )

Arguments

Y_name	Name of the target variable. Character vector of one element.
X_names	Attribute names used for target (Y_name) modelling. Character vector of many elements.
data	Data.frame in which to interpret the parameters Yname and Xnames.
depth	Set the maximum depth of any node of the final tree, with the root node counted as depth 0. Numeric vector of one element which is greater or equal to 0.
min_obs	The minimum number of observations that must exist in any terminal node (leaf). Numeric vector of one element which is greater or equal to 1.
type	Method used for learning. Character vector of one element with one of the: "Shannon", "Renyi", "Tsallis", "Sharma-Mittal", ""Sharma-Taneja", "Kapur".
entropy_par	Numeric vector specifying parameters for the following entropies: "Renyi", "Tsallis", "Sharma-Mittal", "Sharma-Taneja", "Kapur". For "Renyi", "Tsallis" it is one-element vector with q-value. For "Sharma-Mittal" or "Sharma-Taneja" and "Kapura" it is two-element vector with either q-value and r-value or alpha-value and beta-value, respectively.
cp	Complexity parameter, i.e. any split that does not decrease the overall lack of fit by a factor of cp is not attempted. It refers to miss-classification error. If cost or weights are specified aforementioned measure takes these parameter into account. Numeric vector of one element which is greater or equal to 0.
n_cores	Number of cores used for parallel processing. Numeric vector of one element which is greater or equal to 1.
weights	Numeric vector of cases weights. It should have as many elements as the number of observation in the data.frame passed to the data parameter.
cost	Matrix of costs associated with the possible errors. The matrix should have k columns and rows, where k is the number of class levels. Rows contain true classes while columns contain predicted classes. Rows and columns names should take all possible categories (labels) of the target variable.
class_th	Method used for determining thresholds based on which the final class for each node is derived. If cost is specified it can take one of the following: "theoretical", "tuned", otherwise it takes "equal". Character vector of one element.
overfit	Character vector of one element with one of the: "none",”leafcut”, "prune", "avoid" specifying which method overcoming overfitting should be used. ”leafcut” method is used when the full tree is built, it reduces the subtree when both siblings choose the same class label. "avoid" method is incorporated during the recursive partitioning, it prohibit the split when both sibling chose the same class. “prune” method employs pessimistic error pruning procedure, it should be specified along with the cf parameter.
cf	Numeric vector of one element with the number in (0, 1) for the optional pessimistic-error-rate-based pruning step.
amb_prob	Ambiguity threshold for the difference between the highest class probability and the second highest class probability per node, below which the expert has to make a decision regarding the future tree structure. Logical vector with one element. It works when the amb_class parameter is NULL.
top_split	Number of best splits, i.e. final trees structure to be presented. Splits are sorted in descending order according to the information gain. Numeric vector with one element.
var_lev	Decision indicating whether possible best splits are derived on the attribute level (higher) or on the split point for each attribute (lower). “TRUE” means that the expert gets the best splits, one for each variable. "FALSE” means the best splits at all where it might happen that the expert receives top_split splits coming from only one variable. Logical vector with one element.
amb_class	Labels of class for which the expert will make a decision during the learning. Character vector of many elements (from 1 up to number of classes). Should have the same number of elements as vector passed to the amb_class_freq parameter.
amb_class_freq	Classes frequencies per node above which the expert will make a decision. Numeric vector of many elements (from 1 up to number of classes). Should have the same number of elements as vector passed to the amb_class parameter.
tree_path	Path to the folder where the proposed trees created during the interactive learning will be stored. *.txt file with the tree structure is iteratively updated. Character vector with one element.

Value

A fitted model/object of class Node, R6. See. data.tree.

See Also

ImbTreeEntropy, ImbTreeEntropyInter, PredictTree, PrintTree, PrintTreeInter, ExtractRules

Examples

library("ImbTreeEntropy")
data(iris)
# Choosing sequence: 4, 3, 2, 1, 1
Tree <- ImbTreeEntropyInter(Y_name = "Species", 
                            X_names = colnames(iris)[-ncol(iris)], 
                            data = iris) 
PrintTreeInter(Tree)

KrzyGajow/ImbTreeEntropy documentation built on Dec. 31, 2020, 2:13 p.m.