medusagit / rmoa

Connect R to MOA for massive online data stream mining

RMOA

RMOA allows to interface R with MOA (http://moa.cms.waikato.ac.nz/).

RMOA interfaces with MOA version 2014.04. Documentation of MOA directed towards RMOA users can be found at http://jwijffels.github.io/RMOA/

Models

Currently RMOA focusses on classification models (as the stream package in R already allows clustering). Classification models which are possible through RMOA are:

• Classification trees:
  • AdaHoeffdingOptionTree
  • ASHoeffdingTree
  • DecisionStump
  • HoeffdingAdaptiveTree
  • HoeffdingOptionTree
  • HoeffdingTree
  • LimAttHoeffdingTree
  • RandomHoeffdingTree
• Bayesian classification:
  • NaiveBayes
  • NaiveBayesMultinomial
• Active learning classification:
  • ActiveClassifier
• Ensemble (meta) classifiers:
  • Bagging
    • LeveragingBag
    • OzaBag
    • OzaBagAdwin
    • OzaBagASHT
  • Boosting
    • OCBoost
    • OzaBoost
    • OzaBoostAdwin
  • Stacking
    • LimAttClassifier
  • Other
    • AccuracyUpdatedEnsemble
    • AccuracyWeightedEnsemble
    • ADACC
    • DACC
    • OnlineAccuracyUpdatedEnsemble
    • TemporallyAugmentedClassifier
    • WeightedMajorityAlgorithm

Streaming regression models are also included namely

• Rules:
  • TargetMean and FadingTargetMean
  • Perceptron
  • AMRulesRegressor
• Trees:
  • FIMTDD
  • ORTO
• Functions
  • SGD (Stochastic gradient descent)

Streaming recommendation engines which are made available are

• BaselinePredictor
• BRISMFPredictor

Installation

The package is currently available at CRAN.

To install the latest development version from github

library(devtools)
install.packages("ff")
install.packages("rJava")
install_github("jwijffels/RMOA", subdir="RMOAjars/pkg")
install_github("jwijffels/RMOA", subdir="RMOA/pkg")

Examples

Examples below show how to construct, train and score using a HoeffdingTree and boosted ensemble of HoeffdingTree.

require(RMOA)

## Create a HoeffdingTree
hdt <- HoeffdingTree(numericEstimator = "GaussianNumericAttributeClassObserver")
hdt

## Define a stream - e.g. a stream based on a data.frame
data(iris)
iris <- factorise(iris)
irisdatastream <- datastream_dataframe(data=iris)

## Train the HoeffdingTree on the iris dataset
mymodel <- trainMOA(model = hdt, 
  formula = Species ~ Sepal.Length + Sepal.Width + Petal.Length, 
  data = irisdatastream)

## Predict using the HoeffdingTree on the iris dataset
scores <- predict(mymodel, newdata=iris, type="response")
str(scores)
table(scores, iris$Species)
scores <- predict(mymodel, newdata=iris, type="votes")
head(scores)


##
## Boosting example
##
irisdatastream$reset()
mymodel <- OzaBoost(baseLearner = "trees.HoeffdingTree", ensembleSize = 30)
mymodel <- trainMOA(model = mymodel, 
  formula = Species ~ Sepal.Length + Sepal.Width + Petal.Length, 
  data = irisdatastream)

## Predict using the HoeffdingTree on the iris dataset
scores <- predict(mymodel, newdata=iris, type="response")
str(scores)
table(scores, iris$Species)
scores <- predict(mymodel, newdata=iris, type="votes")
head(scores)


##
## Streaming regressions and streaming recommendation engines. Examples can be found in the documentation
##
?trainMOA.MOA_regressor
?trainMOA.MOA_recommender

Streams

Data streams are implemented for classic data in R (data.frame, matrix), data in files (csv, delimited, flat table) as well as out-of memory data in an ffdf (ff package).

TODO

Currently the following MOA models are not (yet) implemented in RMOA.

• Multilabel, drift, functions, rules classifiers
• Outlier detection
• Clustering