regressor: Regression Based on Association Rules

In ianjjohnson/arulesRegression: Regression Based on Association Rules

Description

Build a regression model using a transaction boosting classification by association algorithm. The algorithm is currently in development, and is not yet formally documented.

Usage

bCBA.regressor(formula, data, support = 0.2,
    confidence = 0.8, gamma = 0.05, cost = 10.0,
    num_levels = 20, verbose=FALSE, parameter = NULL,
    control = NULL, sort.parameter=NULL, lhs.support=TRUE,
		disc.method="cluster", disc.categories=10)

Arguments

formula	A symbolic description of the model to be fitted. Has to be of form class ~ .. The class is the variable name (part of the item label before =).
data	A data.frame containing the training data.
support, confidence	Minimum support and confidence for creating association rules.
gamma, cost	Hyperparameters for the bCBA algorithm.
num_levels	The number of levels into which the target will be quantized for classification.
verbose	Optional logical flag to allow verbose execution, where additional intermediary execution information is printed at runtime.
parameter, control	Optional parameter and control lists for apriori.
sort.parameter	Ordered vector of arules interest measures (as characters) which are used to sort rules in preprocessing.
lhs.support	Logical variable, which, when set to default value of True, indicates that LHS support should be used for rule mining.
disc.method	Discretization method for arules discretize function for factorizing numeric input (default: "cluster")
disc.categories	Number of discretization categories for arules discretize function for factorizing numeric input (default: 10)

Details

Uses the bCBA algorithm from arulesCBA to build a transaction-boosted classifier with weighted rules for predicting the target variable. The weighted rules are used in regression by calculating the weighted sum of the means of the class values for the rules which match the new item.

Value

Returns an object of class CBA representing the trained classifier with fields:

rules	the classifier rule base.
default	deault class label.
levels	levels of the class variable.
means	the means of each discretized bucket of the target variable.

Author(s)

Ian Johnson

See Also

predict.CBA, bCBA, apriori, rules, transactions.

Examples

  library(arulesCBA)

  data(iris)
  irisDisc <- as.data.frame(lapply(iris[2:4], function(x) discretize(x, categories=9)))
  irisDisc$Sepal.Length <- iris$Sepal.Length
  irisDisc$Species <- iris$Species

  regressor <- bCBA.regression(Sepal.Length ~ ., irisDisc, support = .01)
result <- predict(regressor, irisDisc)

  #Mean squared error
  sum((result - iris$Sepal.Length)^2) / length(result)
  summary(result)
  summary(iris$Sepal.Length)

ianjjohnson/arulesRegression documentation built on May 18, 2019, 1:28 a.m.