randomForest.ddR (version 0.1.0)

drandomForest: Distributed randomForest with parallelism in sub-forest level

Description

drandomForest function runs randomForest function of randomForest package in a distributed fashion with parallelism in sub-forest level. drandomForest calls several instances of randomForest distributed across a cluster system in order to create sub-forests concurrently. Therefore, the master distributes the input data among all R-executors of the distributedR environment, and trees on different sub-sections of the forest are created simultaneously. At the end, all these trees are combined to result a single forest.

The interface of drandomForest is similar to randomForest. Indeed it adds two arguments nExecutor and trace, and removes several other arguments: subset, do.trace, corr.bias, keep.inbag, and oob.prox. Nevertheless, it must be noticed that default value of some arguments are changed as well to make the algorithm more scalable for big data problems; e.g, proximity is FALSE by default. Its returned result is also completely compatible to the result of randomForest.

Usage

## S3 method for class 'formula':
drandomForest(formula, data=NULL, ..., ntree=500, 
         na.action=na.fail, nExecutor, trace=FALSE, 
         completeModel=FALSE, setSeed)
## S3 method for class 'default':
drandomForest(x, y=NULL,  xtest=NULL, ytest=NULL, ntree=500,
         mtry=if (!is.null(y) && !is.factor(y) && !is.dframe(y))
         max(floor(ncol(x)/3), 1) else floor(sqrt(ncol(x))),
         replace=TRUE, classwt=NULL, cutoff, strata,
         sampsize = if (replace) nrow(x) else ceiling(.632*nrow(x)),
         nodesize = if (!is.null(y) && !is.factor(y) && 
         !is.dframe(y)) 5 else 1,
         maxnodes=NULL, importance=FALSE, localImp=FALSE, nPerm=1,
         proximity=FALSE,norm.votes=TRUE, keep.forest=TRUE,
         nExecutor, trace=FALSE, completeModel=FALSE, ...,
	 setSeed, formula, na.action = na.fail)

Arguments

data

a data frame or dframe which contains samples.

na.action

A function to specify the action to be taken if NAs are found. (NOTE: If given, this argument must be named.)

formula

a formula describing the model to be fitted. It must be a simple formula without any arithmetic operation among columns.

x

when a data frame or a matrix of predictors assigned to x, its size should not be bigger than 2GB. For bigger datasets, darray should be used. darray does not support categorical data. Therefore, dframe and the first interface should be used

y

a response vector. If a factor, classification is assumed, otherwise regression is assumed. If omitted, hpdRF_parallelForest will run in unsupervised mode. When x is a darray), y should be also a darray

xtest

a data frame or matrix (like x) containing predictors for the test set. When x is a darray, it should be of the same type.

ytest

response for the test set. Its type should be consistent with y. Moreover, it should have a single column.

ntree

Number of trees to grow. This should not be set to too small a number, to ensure that every input row gets predicted at least a few times.

mtry

Number of variables randomly sampled as candidates at each split. Note that the default values are different for classification (sqrt(p) where p is number of variables in x) and regression (p/3)

replace

Should sampling of cases be done with or without replacement?

classwt

Priors of the classes. Need not add up to one. Ignored for regression.

cutoff

(Classification only) A vector of length equal to number of classes. The 'winning' class for an observation is the one with the maximum ratio of proportion of votes to cutoff. Default is 1/k where k is the number of classes (i.e., majority v

strata

A (factor) variable that is used for stratified sampling.

sampsize

Size(s) of sample to draw. For classification, if sampsize is a vector of the length the number of strata, then sampling is stratified by strata, and the elements of sampsize indicate the numbers to be drawn from the strata.

nodesize

Minimum size of terminal nodes. Setting this number larger causes smaller trees to be grown (and thus take less time). Note that the default values are different for classification (1) and regression (5).

maxnodes

Maximum number of terminal nodes trees in the forest can have. If not given, trees are grown to the maximum possible (subject to limits by nodesize). If set larger than maximum possible, a warning is issued.

importance

Should importance of predictors be assessed?

localImp

Should casewise importance measure be computed? (Setting this to TRUE will override importance.)

nPerm

Number of times the OOB data are permuted per tree for assessing variable importance. Number larger than 1 gives slightly more stable estimate, but not very effective. Currently only implemented for regression.

proximity

a logical value which indicates if the proximity measure among the rows should be calculated. It is FALSE by default because it is very memory inefficient. Moreover, it is calculated only on 'out-of-bag' data

norm.votes

If TRUE (default), the final result of votes are expressed as fractions. If FALSE, raw vote counts are returned (useful for combining results from different runs). Ignored for regression.

keep.forest

If set to FALSE, the forest will not be retained in the output object.

nExecutor

a positive integer number indicating the number of tasks for running the function. To have optimal performance, it is recommended to have this number smaller than or equal to the total number of cores.

trace

when this argument is true, intermediate steps of the progress are displayed.

completeModel

when it is FALSE (default), the output values that preserve information per sample are discarded. They are 'oob.times', 'votes', 'predicted', 'confusion', 'err.rate', 'mse', 'rsq', 'proximity', and 'test'. This feature is intended to keep the size of the

...

optional parameters to be passed to the low level function.

setSeed

if setSeed is a valid integer, the output model will be deterministic assuming nExecutor parameter does not change

Value

• An object of class randomForest. The result is similar to the result of the combine function in randomForest package and will contain the following components.
• call
  the original call to drandomForest
• type
  one of regression, classification, or unsupervised.
• predicted
  (only when completeModel=TRUE) the predicted values of the input data based on out-of-bag samples.
• importance
  a matrix with nclass + 2 (for classification) or two (for regression) columns. For classification, the first nclass columns are the class-specific measures computed as mean descrease in accuracy. The nclass + 1st column is the mean descrease in accuracy over all classes. The last column is the mean decrease in Gini index. For Regression, the first column is the mean decrease in accuracy and the second the mean decrease in MSE. If importance=FALSE, the last measure is still returned as a vector.
• importanceSD
  The ''standard errors'' of the permutation-based importance measure. For classification, a p by nclass + 1 matrix corresponding to the first nclass + 1 columns of the importance matrix. For regression, a length p vector.
• localImp
  a p by n matrix containing the casewise importance measures, the [i,j] element of which is the importance of i-th variable on the j-th case. NULL if localImp=FALSE.
• ntree
  number of trees grown.
• mtry
  number of predictors sampled for spliting at each node.
• forest
  (a list that contains the entire forest; NULL if hpdRF_parallelForest is run in unsupervised mode or if keep.forest=FALSE.
• err.rate
  (classification only) vector error rates of the prediction on the input data, the i-th element being the (OOB) error rate for all trees up to the i-th.
• confusion
  (classification only) the confusion matrix of the prediction (based on OOB data).
• votes
  (classification only, and only when completeModel=TRUE) a matrix with one row for each input data point and one column for each class, giving the fraction or number of (OOB) 'votes' from the random forest.
• oob.times
  (only when completeModel=TRUE) number of times cases are 'out-of-bag' (and thus used in computing OOB error estimate)
• y
  (only when completeModel=TRUE) the response vector if it is made available in the input.
• proximity
  (only when completeModel=TRUE) a matrix of proximity measures among the input (based on the frequency that pairs of data points are in the same terminal nodes) when proximity=TRUE.
• mse
  (regression only) vector of mean square errors: sum of squared residuals divided by n.
• rsq
  (regression only) ''pseudo R-squared'': 1 - mse / Var(y).
• test
  (only when completeModel=TRUE) if test set is given (through the xtest or additionally ytest arguments), this component is a list which contains the corresponding predicted, err.rate, confusion, votes (for classification) or predicted, mse and rsq (for regression) for the test set.
• terms
  it contains a formula identifying response and predictors (for classification and regression types).

References

Breiman, L. (2001), Random Forests, Machine Learning 45(1),5-32.

Random Forests V4.6-10, http://cran.r-project.org/web/packages/randomForest/randomForest.pdf.

Examples

library(ddR.randomForest)

## Classification:
##data(iris)
iris.rf <- drandomForest(Species ~ ., data=iris, importance=TRUE)
print(iris.rf)

## The 'unsupervised' case:
iris.urf <- drandomForest(iris[, -5], 
                            proximity=TRUE, completeModel=TRUE)
MDSplot(iris.urf, iris$Species)

## stratified sampling: draw 20, 30, and 20 of the species to grow each tree.
(iris.rf2 <- drandomForest(iris[1:4], iris$Species, 
                          sampsize=c(20, 30, 20)))

## Regression:
## data(airquality)
ozone.rf <- drandomForest(Ozone ~ ., data=airquality, mtry=3,
                    importance=TRUE, na.action=na.omit, 
                    completeModel=TRUE)
print(ozone.rf)
## Show "importance" of variables: higher value mean more important:
round(importance(ozone.rf), 2)

## "x" can be a matrix instead of a data frame:
x <- matrix(runif(5e2), 100)
y <- gl(2, 50)
(myrf <- drandomForest(x, y))
(predict(myrf, x))

## "complicated" formula:
(swiss.rf <- drandomForest(sqrt(Fertility)~. - Catholic + I(Catholic<50),
                          data=swiss))
(predict(swiss.rf, swiss))
## Test use of 32-level factor as a predictor:
x <- data.frame(x1=gl(32, 5), x2=runif(160), y=rnorm(160))
(rf1 <- drandomForest(x[-3], x[[3]], ntree=10))

## Grow no more than 4 nodes per tree:
(treesize(drandomForest(Species ~ ., data=iris, maxnodes=4, ntree=30)))

distributedR_shutdown()