README.md

In kliegr/marc: Title: rMARC

Monotonicity Exploiting Association Rule Classification

This package for R implements the MARC algorithm aimed at postprocesing output of a CBA classifier.

Liu, B. Hsu, W. and Ma, Y (1998). Integrating Classification and Association Rule Mining. Proceedings KDD-98, New York, 27-31 August. AAAI. pp 80-86.

The arc package is used for generation of the CBA classifier.

Features

• Java implementation with R wrapper
• Supports one rule as well as multi rule classification
• Optional fuzzification of rules
• Optional annotation of rules with probability distributions
• Optional Post pruning -- less aggressive, generally slightly decreases accuracy as well as rule count
• Optional Continuous pruning -- more aggressive, generally higher decrease in accuracy as well as rule count

Installation

Package can be installed from the R environment using the devtools package.

devtools::install_github("kliegr/marc")

Examples

One rule classification

  library(rMARC)
  library(mlbench)
  data("PimaIndiansDiabetes")

  set.seed(111)
  allData <- PimaIndiansDiabetes[sample(nrow(PimaIndiansDiabetes)),]
  trainFold <- allData[1:500,]
  testFold <- allData[501:nrow(PimaIndiansDiabetes),]
  rmCBA <- cba(trainFold, classAtt="diabetes")
  prediction <- predict(rmCBA,testFold)
  acc <- CBARuleModelAccuracy(prediction, testFold[[rmCBA@classAtt]])
  print(paste("CBA Model with ",length(rmCBA@rules), " rules and accuracy ",acc))
  rmMARC <- marcExtend(cbaRuleModel=rmCBA,datadf=trainFold,continuousPruning=FALSE, postpruning=TRUE, fuzzification=FALSE, annotate=FALSE,ruleOutputPath="rules.xml")
  prediction <- predict(rmMARC,testFold,"oneRule")
  acc <- CBARuleModelAccuracy(prediction, testFold[[rmMARC@classAtt]])
  print(paste("MARC Model with ",rmMARC@ruleCount, " rules and accuracy ",acc))
  print(rmMARC@rules)

Output

[1] CBA Model with  53  rules and accuracy  0.753731343283582
[1] MARC Model with  47  rules and accuracy  0.753731343283582

MARC decreased the number of rules while keeping same accuracy.

If we actived the continuousPruning option, it would result in aggressive pruning:

[1] MARC Model with  28  rules and accuracy  0.67910447761194

Multi rule classification

  library(rMARC)
  library(mlbench)
  data("Ionosphere")
  library(rMARC)
  set.seed(111)
  allData <- Ionosphere[sample(nrow(Ionosphere)),]
  trainFold <- allData[1:300,]
  testFold <- allData[301:nrow(Ionosphere),]
  rmCBA <- cba(trainFold, classAtt="Class")
  prediction <- predict(rmCBA,testFold)
  acc <- CBARuleModelAccuracy(prediction, testFold[[rmCBA@classAtt]])
  print(paste("CBA Model with ",length(rmCBA@rules), " rules and accuracy ",acc))
  rmMARC <- marcExtend(cbaRuleModel=rmCBA,datadf=trainFold,continuousPruning=TRUE, postpruning=TRUE, fuzzification=FALSE, annotate=TRUE,ruleOutputPath="rules.xml")
  prediction <- predict(rmMARC,testFold,"mixture")
  acc <- CBARuleModelAccuracy(prediction, testFold[[rmMARC@classAtt]])
  print(paste("MARC Model with ",rmMARC@ruleCount, " rules and accuracy ",acc))
  print(rmMARC@rulePath)

Output

[1] CBA Model with  14  rules and accuracy  0.941176470588235
[1] MARC Model with  14  rules and accuracy  0.96078431372549

MARC improved accuracy while keeping the number of rules. Rules in MARC multi rule model cannot be currently visualized - they are stored in a file.

Evaluation

A lightweight benchmarking framework for MARC is available as marcbench.

kliegr/marc documentation built on May 20, 2019, 12:33 p.m.