rfsrc.anonymous: Anonymous Random Forests

View source: R/rfsrc.anonymous.R

rfsrc.anonymousR Documentation

Anonymous Random Forests

Description

Anonymous random forests applies random forests but is carefully modified so as not to save the original training data. This allows users to share their forest with other researchers but without having to share their original data.

Usage

rfsrc.anonymous(formula, data, forest = TRUE, ...)

Arguments

formula

A symbolic description of the model to be fit. If missing, unsupervised splitting is implemented.

data

Data frame containing the y-outcome and x-variables.

forest

Should the forest object be returned? Used for prediction on new data and required by many of the package functions.

...

Further arguments as in rfsrc. See the rfsrc help file for details.

Details

Calls rfsrc and returns an object with the training data removed so that users can share their forest while maintaining privacy of their data.

In order to predict on test data, it is however necessary for certain minimal information to be saved from the training data. This includes the names of the original variables, and if factor variables are present, the levels of the factors. The mean value and maximal class value for real and factor variables in the training data are also stored for the purposes of imputation on test data (see below). The topology of grow trees is also saved, which includes among other things, the split values used for splitting tree nodes.

For the most privacy, we recommend that variable names be made non-identifiable and that data be coerced to real values. If factors are required, the user should consider using non-identifiable factor levels. However, in all cases, it is the users responsibility to de-identify their data and to check that data privacy holds. We provide NO GUARANTEES of this.

Missing data is especially delicate with anonymous forests. Training data cannot be imputed and the option na.action="na.impute" simply reverts to na.action="na.omit". Therefore if you have training data with missing values consider using pre-imputing the data using impute. It is however possible to impute on test data. The option na.action="na.impute" in the prediction call triggers a rough and fast imputation method where the value of missing test data are replaced by the mean (or maximal class) value from the training data. A second option na.action="na.random" uses a fast random imputation method.

In general, it is important to keep in mind that while anonymous forests tries to play nice with other functions in the package, it only works with calls that do not specifically require training data.

Value

An object of class (rfsrc, grow, anonymous).

Author(s)

Hemant Ishwaran and Udaya B. Kogalur

See Also

rfsrc

Examples



## ------------------------------------------------------------
## regression
## ------------------------------------------------------------
print(rfsrc.anonymous(mpg ~ ., mtcars))

## ------------------------------------------------------------
## plot anonymous regression tree (using get.tree)
## TBD CURRENTLY NOT IMPLEMENTED 
## ------------------------------------------------------------
## plot(get.tree(rfsrc.anonymous(mpg ~ ., mtcars), 10))

## ------------------------------------------------------------
## classification
## ------------------------------------------------------------
print(rfsrc.anonymous(Species ~ ., iris))

## ------------------------------------------------------------
## survival
## ------------------------------------------------------------
data(veteran, package = "randomForestSRC")
print(rfsrc.anonymous(Surv(time, status) ~ ., data = veteran))

## ------------------------------------------------------------
## competing risks
## ------------------------------------------------------------
data(wihs, package = "randomForestSRC")
print(rfsrc.anonymous(Surv(time, status) ~ ., wihs, ntree = 100))

## ------------------------------------------------------------
## unsupervised forests
## ------------------------------------------------------------
print(rfsrc.anonymous(data = iris))

## ------------------------------------------------------------
## multivariate regression
## ------------------------------------------------------------
print(rfsrc.anonymous(Multivar(mpg, cyl) ~., data = mtcars))

## ------------------------------------------------------------
## prediction on test data with missing values using pbc data
## cases 1 to 312 have no missing values
## cases 313 to 418 having missing values
## ------------------------------------------------------------
data(pbc, package = "randomForestSRC")
pbc.obj <- rfsrc.anonymous(Surv(days, status) ~ ., pbc)
print(pbc.obj)

## mean value imputation
print(predict(pbc.obj, pbc[-(1:312),], na.action = "na.impute"))

## random imputation
print(predict(pbc.obj, pbc[-(1:312),], na.action = "na.random"))

## ------------------------------------------------------------
## train/test setting but tricky because factor labels differ over
## training and test data
## ------------------------------------------------------------

# first we convert all x-variables to factors
data(veteran, package = "randomForestSRC")
veteran.factor <- data.frame(lapply(veteran, factor))
veteran.factor$time <- veteran$time
veteran.factor$status <- veteran$status

# split the data into train/test data (25/75)
# the train/test data have the same levels, but different labels
train <- sample(1:nrow(veteran), round(nrow(veteran) * .5))
summary(veteran.factor[train, ])
summary(veteran.factor[-train, ])

# grow the forest on the training data and predict on the test data
v.grow <- rfsrc.anonymous(Surv(time, status) ~ ., veteran.factor[train, ]) 
v.pred <- predict(v.grow, veteran.factor[-train, ])
print(v.grow)
print(v.pred)




randomForestSRC documentation built on Sept. 11, 2024, 7:50 p.m.