arulesCBA (version 1.0)

CBA: Classification Based on Association Rules

Description

Build a classifier based on association rules mined for an input dataset. The CBA algorithm used is a modified version of the algorithm described by Liu, et al. 1998.

Usage

CBA(data, class, support = 0.2, confidence = 0.8, verbose=FALSE)

Arguments

data

A data.frame or an object of class transactions (from arules) containing the training data.

class

A character vector of size 1 indicating the name of the variable which represents the class of the input data entries.

support

The minimum support for association rules mined from apriori

confidence

The minimum confidence for association rules mined from apriori

verbose

Optional logical flag to allow verbose execution, where additional intermediary execution information is printed at runtime.

Value

Returns an object of class CBA representing the trained classifier.

Details

Formats the input data frame and calls a C implementation of the CBA algorithm from Liu, et al. 1998 split up into three stages to build a classifier based on a set of association rules.

Before the CBA algorithm in C is executed, association rules are generated with Apriori from the arules package.

A default class is selected for the classifier. Note that for datasets which do not yield any strong association rules it's possible that no rules will be included in the classifier, and only a default class.

Examples

# prepare data
data(iris)
irisDisc <- as.data.frame(lapply(iris[1:4], function(x) discretize(x, categories=9)))
irisDisc$Species <- iris$Species

classifier <- CBA(irisDisc, "Species", supp = 0.05, conf=0.9)
classifier

# training a classifier with transaction data (arules)
trans <- as(irisDisc, "transactions")
trans
classifier <- CBA(trans, "Species", supp = 0.05, conf=0.9)
classifier