R/generateStream.R
In edouardfouche/R-streamgenerator: Statistical multidimensional synthetic stream generator
#' Generate Static Data Stream
#'
#' Generate a new synthetic multidimensional static data stream having the
#' desired properties.
#'
#' @param n A vector containing \code{x} values, where the values corresponds
#' to the number of points for each step and \code{x} to the number of
#' steps.
#' @param prop Proportion of outliers in the hidden space.
#' @param proptype Type of the proportion of outliers. Value "proportional":
#' depend on the size of the empty space. Value "absolute": same absolute
#' proportion per subspace.
#' @param stream.config A stream configuration object. Should have been
#' generated with \code{nstep = 1}.
#'
#' @return An object of class stream, which is a List of 5 elements.
#' - \code{data} contains the stream generated
#' - \code{labels} contains the description of each point (\code{0} if the point
#' is not an outlier, or the subspace in which it is outlying as a string)
#' - \code{n} the number of points at each step
#' - \code{prop} the proportion of outliers in the hidden space
#' - \code{stream.config} the associated stream configuration object (which is
#' valid only for static streams)
#'
#' @details
#' The data is generated uniformly, except in certain subspaces where the data
#' is concentrated in particular dependencies (i.e. in the "Wall" dependency,
#' data concentrates on the axes, in a L-like shape). This should create spaces
#' with high dependency and space to observe hidden outliers. Note that the
#' proportion of outlier \code{prop} does not relate directly to the percentage
#' of outliers in the output stream. Since it corresponds to the probability of
#' a point, being ALREADY in the hidden space to stay where it is, the overall
#' proportion of outliers depends on the hidden space volume, which depends
#' on the number of subspaces and their margins. The greater the margin, the
#' bigger the hidden space.
#'
#' @examples
#' # Generate a stream with default parameters
#' stream <- generate.static.stream()
#' # Generate a stream with custom configuration
#' stream.config <- generate.stream.config(dim=50, nstep=1) # nstep should be 1
#' stream <- generate.static.stream(n=1000, prop=0.05,
#' stream.config=stream.config)
#' # Output stream results (to uncomment)
#' # output.stream(stream, "example")
#'
#' @author Edouard Fouché, \email{edouard.fouche@kit.edu}
#'
#' @seealso
#' * \code{\link{generate.stream.config}} : generate a stream.config file for a
#' dynamic or static stream
#'
#' @md
#' @export
generate.static.stream <- function(n=1000, prop=0.01, proptype="proportional",
stream.config=NULL) {
# Generate n points with dim dimensions where the list of subspaces are
# generated wall-like with the size of the wall taken from margins list as
# 1-margin. In the hidden space, a proportion prop of the points is taken as
# outliers. Suggestion: add a verbose mode
sanitycheck.generate(n=n, prop=prop, stream.config=stream.config)
if(is.null(stream.config)) {
stream.config <- generate.stream.config(nstep=1)
} else {
if(stream.config$nstep != 1) {
stop("The stream.config file is not compatible with static streams:" +
" nstep should be = 1")
}
}
dim <- stream.config$dim
subspaces <- stream.config$subspaces
margins <- stream.config$margins
dependency <- stream.config$dependency
discretize <- stream.config$discretize
allowOverlap <- stream.config$allowOverlap
meta <- generate.multiple.rows(n, dim, subspaces, margins, prop,
proptype=proptype, dependency=dependency,
discretize=discretize)
res <- list("data"=meta$data,"labels"=meta$labels, "n"=n, "prop"=prop,
"proptype"=proptype, "allowOverlap"=allowOverlap,
"stream.config"=stream.config)
attr(res, "class") <- "stream"
return(res)
}
#' Generate Dynamic Data Stream
#'
#' Generate a new synthetic multidimensional dynamic data stream having the
#' desired properties.
#'
#' @param n A vector containing \code{x} values, where the values corresponds to
#' the number of points for each step and \code{x} to the number of steps.
#' @param prop Proportion of outliers in the hidden space.
#' @param proptype Type of the proportion of outliers. Value "proportional":
#' depend on the size of the empty space. Value "absolute": same absolute
#' proportion per subspace.
#' @param stream.config A stream configuration object. Should have been
#' generated with \code{nstep > 1}.
#' @param verbose If TRUE, then the state of the stream will be printed as
#' output for every 100 points.
#' @param coldstart If TRUE (default) all subspaces will start with a margin
#' value of 0.
#' @param transition A string indication what kind of transition should occur.
#' Can be "Linear" (default) or "Abrupt".
#'
#' @return A an object of class \code{stream}, which is a \code{List} of 5
#' elements.
#' - \code{data} contains the stream generated
#' - \code{labels} contains the description of each point (\code{0} if the point
#' is not an outlier, or the subspace in which it is outlying as a string)
#' - \code{n} the number of points at each step
#' - \code{prop} the proportion of outliers in the hidden space
#' - \code{stream.config} the associated stream configuration object (which is
#' valid only for dynamic streams)
#'
#' @details
#' The data is generated uniformly, except in certain subspaces where the data
#' is concentrated in particular dependencies (i.e. in the "Wall" dependency,
#' data concentrates on the axes, in a L-like shape). This should create spaces
#' with high dependency and space to observe hidden outliers. Note that the
#' proportion of outlier \code{prop} does not relate directly to the percentage
#' of outliers in the output stream. Since it corresponds to the probability of
#' a point, being ALREADY in the hidden space to stay where it is, the overall
#' proportion of outliers depends on the hidden space volume, which depends
#' on the number of subspaces and their margins. The greater the margin, the
#' bigger the hidden space.
#'
#' @examples
#' # Generate a stream with default parameters
#' stream <- generate.dynamic.stream()
#' # Generate a stream with custom configuration
#' stream.config <- generate.stream.config(dim=50, nstep=10, volatility=0.5)
#' stream <- generate.dynamic.stream(n=100, prop=0.05,
#' stream.config=stream.config)
#' # Output stream results (to uncomment)
#' # output.stream(stream, "example")
#'
#' @author Edouard Fouché, \email{edouard.fouche@kit.edu}
#'
#' @seealso
#' * \code{\link{generate.stream.config}} : generate a stream.config file for a
#' dynamic or static stream
#'
#' @md
#' @export
generate.dynamic.stream <- function(n=100, prop=0.01, proptype="proportional",
stream.config=NULL, verbose=FALSE,
coldstart=TRUE, transition="Linear") {
sanitycheck.generate(n=n, prop=prop, stream.config=stream.config,
verbose=verbose)
if(is.null(stream.config)) {
stream.config <- generate.stream.config()
} else {
if(stream.config$nstep <= 1) {
stop("The stream.config file in not compatible with dynamic streams:" +
" nstep should be > 1")
}
}
if(length(n) == 1) {
n <- rep(n, stream.config$nstep)
} # else assume that n has the good size, was checked by the sanity check
dim <- stream.config$dim
subspaceslist <- stream.config$subspaceslist
marginslist <- stream.config$marginslist
dependency <- stream.config$dependency
discretize <- stream.config$discretize
allowOverlap <- stream.config$allowOverlap
data <- data.frame()
labels <- c()
# Generate some data for each time step description
for(seq in 1:length(n)) {
if(verbose) print(paste("Step", seq, "of", length(n), ". Size", n[[seq]],
"elements."))
# Determine for the current state step the start and end margins values for
# each subspaces
subspaces_state <- list() # Indicates if the subspace has a dependency.
currentmargins <- list() # The start margin-value of a step.
nextmargins <- list() # The end margin-value of a step.
if(seq == 1) {
# If we want a coldstart, the starting margins values will be 0 for all
# subspaces. Otherwise, the provided value is used.
# TODO @apoth: Check, if this influences whether a drift is possible in
# from the first to the second step.
subspaces_state <- subspaceslist[[seq]]
if(coldstart) {
currentmargins <- c(rep(0, length(subspaceslist[[seq]])))
} else {
currentmargins <- marginslist[[seq]]
}
nextmargins <- marginslist[[seq]]
} else {
# We shall consider subspace from the previous and the next state
subspaces_state <- unique(c(subspaceslist[[seq - 1]],
subspaceslist[[seq]]))
for(sub in 1:length(subspaces_state)) {
# In the case a subspace is contained in the next step, its intended
# value should be equal to its margins in the next step.
# Otherwise, it should be 0.
if(any(sapply(subspaceslist[[seq]],
function(x) setequal(x, subspaces_state[[sub]])))) {
nextmargins <- c(nextmargins,
marginslist[[seq]][sapply(subspaceslist[[seq]],
function(x) setequal(x, subspaces_state[[sub]]))])
} else {
nextmargins <- c(nextmargins, 0)
}
# In the case a subspace is contained in the current step, its start
# value should be equal to its margins in the current step.
# Otherwise, it should be 0.
if(any(sapply(subspaceslist[[seq - 1]],
function(x) setequal(x,subspaces_state[[sub]])))) {
currentmargins <- c(currentmargins,
marginslist[[seq - 1]][sapply(subspaceslist[[seq - 1]],
function(x) setequal(x,subspaces_state[[sub]]))])
} else {
currentmargins <- c(currentmargins, 0)
}
}
}
currentmargins <- as.list(currentmargins)
nextmargins <- as.list(nextmargins)
i <- 0
for(x in 1:n[[seq]]) {
# TODO @apoth: Add new transition types here!
#
# Update the current margins (transitioning uniformly between
# currentmargins and nextmargins)
if(transition == "Linear") {
margins_state <- as.list(unlist(currentmargins) -
(unlist(currentmargins) -
unlist(nextmargins)) * (x - 1) / n[[seq]])
} else if(transition == "Abrupt") { # Leave the margins as they are
margins_state <- currentmargins
} else {
stop("Unknown transition type specified.")
}
if(i %% 100 == 0 & verbose) {
print(c("subspaces_state:", paste(subspaces_state)), collapse=" ")
#print(c("currentmargins:", paste(currentmargins)), collapse=" ")
print(c("margins_state:", paste(margins_state)), collapse=" ")
#print(c("nextmargins:", paste(nextmargins)), collapse=" ")
}
i <- i + 1
# Generate a row
res <- generate.row(dim=dim, subspaces=subspaces_state,
margins=margins_state, prop=prop, proptype=proptype,
dependency=dependency, discretize=discretize)
data <- rbind(data, t(res$data))
labels <- c(labels, res$label)
}
}
# Put adequate names on the columns
attributes(data)$names <- c(c(1:dim),"class")
res <- list("data"=data,"labels"=labels, "n"=n, "prop"=prop,
"proptype"=proptype, "allowOverlap" = allowOverlap,
"stream.config"=stream.config)
attr(res, "class") <- "stream"
return(res)
}
edouardfouche/R-streamgenerator documentation built on May 15, 2019, 11:02 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
#' Generate Static Data Stream
#'
#' Generate a new synthetic multidimensional static data stream having the
#' desired properties.
#'
#' @param n A vector containing \code{x} values, where the values corresponds
#' to the number of points for each step and \code{x} to the number of
#' steps.
#' @param prop Proportion of outliers in the hidden space.
#' @param proptype Type of the proportion of outliers. Value "proportional":
#' depend on the size of the empty space. Value "absolute": same absolute
#' proportion per subspace.
#' @param stream.config A stream configuration object. Should have been
#' generated with \code{nstep = 1}.
#'
#' @return An object of class stream, which is a List of 5 elements.
#' - \code{data} contains the stream generated
#' - \code{labels} contains the description of each point (\code{0} if the point
#' is not an outlier, or the subspace in which it is outlying as a string)
#' - \code{n} the number of points at each step
#' - \code{prop} the proportion of outliers in the hidden space
#' - \code{stream.config} the associated stream configuration object (which is
#' valid only for static streams)
#'
#' @details
#' The data is generated uniformly, except in certain subspaces where the data
#' is concentrated in particular dependencies (i.e. in the "Wall" dependency,
#' data concentrates on the axes, in a L-like shape). This should create spaces
#' with high dependency and space to observe hidden outliers. Note that the
#' proportion of outlier \code{prop} does not relate directly to the percentage
#' of outliers in the output stream. Since it corresponds to the probability of
#' a point, being ALREADY in the hidden space to stay where it is, the overall
#' proportion of outliers depends on the hidden space volume, which depends
#' on the number of subspaces and their margins. The greater the margin, the
#' bigger the hidden space.
#'
#' @examples
#' # Generate a stream with default parameters
#' stream <- generate.static.stream()
#' # Generate a stream with custom configuration
#' stream.config <- generate.stream.config(dim=50, nstep=1) # nstep should be 1
#' stream <- generate.static.stream(n=1000, prop=0.05,
#' stream.config=stream.config)
#' # Output stream results (to uncomment)
#' # output.stream(stream, "example")
#'
#' @author Edouard Fouché, \email{edouard.fouche@kit.edu}
#'
#' @seealso
#' * \code{\link{generate.stream.config}} : generate a stream.config file for a
#' dynamic or static stream
#'
#' @md
#' @export
generate.static.stream <- function(n=1000, prop=0.01, proptype="proportional",
stream.config=NULL) {
# Generate n points with dim dimensions where the list of subspaces are
# generated wall-like with the size of the wall taken from margins list as
# 1-margin. In the hidden space, a proportion prop of the points is taken as
# outliers. Suggestion: add a verbose mode
sanitycheck.generate(n=n, prop=prop, stream.config=stream.config)
if(is.null(stream.config)) {
stream.config <- generate.stream.config(nstep=1)
} else {
if(stream.config$nstep != 1) {
stop("The stream.config file is not compatible with static streams:" +
" nstep should be = 1")
}
}
dim <- stream.config$dim
subspaces <- stream.config$subspaces
margins <- stream.config$margins
dependency <- stream.config$dependency
discretize <- stream.config$discretize
allowOverlap <- stream.config$allowOverlap
meta <- generate.multiple.rows(n, dim, subspaces, margins, prop,
proptype=proptype, dependency=dependency,
discretize=discretize)
res <- list("data"=meta$data,"labels"=meta$labels, "n"=n, "prop"=prop,
"proptype"=proptype, "allowOverlap"=allowOverlap,
"stream.config"=stream.config)
attr(res, "class") <- "stream"
return(res)
}
#' Generate Dynamic Data Stream
#'
#' Generate a new synthetic multidimensional dynamic data stream having the
#' desired properties.
#'
#' @param n A vector containing \code{x} values, where the values corresponds to
#' the number of points for each step and \code{x} to the number of steps.
#' @param prop Proportion of outliers in the hidden space.
#' @param proptype Type of the proportion of outliers. Value "proportional":
#' depend on the size of the empty space. Value "absolute": same absolute
#' proportion per subspace.
#' @param stream.config A stream configuration object. Should have been
#' generated with \code{nstep > 1}.
#' @param verbose If TRUE, then the state of the stream will be printed as
#' output for every 100 points.
#' @param coldstart If TRUE (default) all subspaces will start with a margin
#' value of 0.
#' @param transition A string indication what kind of transition should occur.
#' Can be "Linear" (default) or "Abrupt".
#'
#' @return A an object of class \code{stream}, which is a \code{List} of 5
#' elements.
#' - \code{data} contains the stream generated
#' - \code{labels} contains the description of each point (\code{0} if the point
#' is not an outlier, or the subspace in which it is outlying as a string)
#' - \code{n} the number of points at each step
#' - \code{prop} the proportion of outliers in the hidden space
#' - \code{stream.config} the associated stream configuration object (which is
#' valid only for dynamic streams)
#'
#' @details
#' The data is generated uniformly, except in certain subspaces where the data
#' is concentrated in particular dependencies (i.e. in the "Wall" dependency,
#' data concentrates on the axes, in a L-like shape). This should create spaces
#' with high dependency and space to observe hidden outliers. Note that the
#' proportion of outlier \code{prop} does not relate directly to the percentage
#' of outliers in the output stream. Since it corresponds to the probability of
#' a point, being ALREADY in the hidden space to stay where it is, the overall
#' proportion of outliers depends on the hidden space volume, which depends
#' on the number of subspaces and their margins. The greater the margin, the
#' bigger the hidden space.
#'
#' @examples
#' # Generate a stream with default parameters
#' stream <- generate.dynamic.stream()
#' # Generate a stream with custom configuration
#' stream.config <- generate.stream.config(dim=50, nstep=10, volatility=0.5)
#' stream <- generate.dynamic.stream(n=100, prop=0.05,
#' stream.config=stream.config)
#' # Output stream results (to uncomment)
#' # output.stream(stream, "example")
#'
#' @author Edouard Fouché, \email{edouard.fouche@kit.edu}
#'
#' @seealso
#' * \code{\link{generate.stream.config}} : generate a stream.config file for a
#' dynamic or static stream
#'
#' @md
#' @export
generate.dynamic.stream <- function(n=100, prop=0.01, proptype="proportional",
stream.config=NULL, verbose=FALSE,
coldstart=TRUE, transition="Linear") {
sanitycheck.generate(n=n, prop=prop, stream.config=stream.config,
verbose=verbose)
if(is.null(stream.config)) {
stream.config <- generate.stream.config()
} else {
if(stream.config$nstep <= 1) {
stop("The stream.config file in not compatible with dynamic streams:" +
" nstep should be > 1")
}
}
if(length(n) == 1) {
n <- rep(n, stream.config$nstep)
} # else assume that n has the good size, was checked by the sanity check
dim <- stream.config$dim
subspaceslist <- stream.config$subspaceslist
marginslist <- stream.config$marginslist
dependency <- stream.config$dependency
discretize <- stream.config$discretize
allowOverlap <- stream.config$allowOverlap
data <- data.frame()
labels <- c()
# Generate some data for each time step description
for(seq in 1:length(n)) {
if(verbose) print(paste("Step", seq, "of", length(n), ". Size", n[[seq]],
"elements."))
# Determine for the current state step the start and end margins values for
# each subspaces
subspaces_state <- list() # Indicates if the subspace has a dependency.
currentmargins <- list() # The start margin-value of a step.
nextmargins <- list() # The end margin-value of a step.
if(seq == 1) {
# If we want a coldstart, the starting margins values will be 0 for all
# subspaces. Otherwise, the provided value is used.
# TODO @apoth: Check, if this influences whether a drift is possible in
# from the first to the second step.
subspaces_state <- subspaceslist[[seq]]
if(coldstart) {
currentmargins <- c(rep(0, length(subspaceslist[[seq]])))
} else {
currentmargins <- marginslist[[seq]]
}
nextmargins <- marginslist[[seq]]
} else {
# We shall consider subspace from the previous and the next state
subspaces_state <- unique(c(subspaceslist[[seq - 1]],
subspaceslist[[seq]]))
for(sub in 1:length(subspaces_state)) {
# In the case a subspace is contained in the next step, its intended
# value should be equal to its margins in the next step.
# Otherwise, it should be 0.
if(any(sapply(subspaceslist[[seq]],
function(x) setequal(x, subspaces_state[[sub]])))) {
nextmargins <- c(nextmargins,
marginslist[[seq]][sapply(subspaceslist[[seq]],
function(x) setequal(x, subspaces_state[[sub]]))])
} else {
nextmargins <- c(nextmargins, 0)
}
# In the case a subspace is contained in the current step, its start
# value should be equal to its margins in the current step.
# Otherwise, it should be 0.
if(any(sapply(subspaceslist[[seq - 1]],
function(x) setequal(x,subspaces_state[[sub]])))) {
currentmargins <- c(currentmargins,
marginslist[[seq - 1]][sapply(subspaceslist[[seq - 1]],
function(x) setequal(x,subspaces_state[[sub]]))])
} else {
currentmargins <- c(currentmargins, 0)
}
}
}
currentmargins <- as.list(currentmargins)
nextmargins <- as.list(nextmargins)
i <- 0
for(x in 1:n[[seq]]) {
# TODO @apoth: Add new transition types here!
#
# Update the current margins (transitioning uniformly between
# currentmargins and nextmargins)
if(transition == "Linear") {
margins_state <- as.list(unlist(currentmargins) -
(unlist(currentmargins) -
unlist(nextmargins)) * (x - 1) / n[[seq]])
} else if(transition == "Abrupt") { # Leave the margins as they are
margins_state <- currentmargins
} else {
stop("Unknown transition type specified.")
}
if(i %% 100 == 0 & verbose) {
print(c("subspaces_state:", paste(subspaces_state)), collapse=" ")
#print(c("currentmargins:", paste(currentmargins)), collapse=" ")
print(c("margins_state:", paste(margins_state)), collapse=" ")
#print(c("nextmargins:", paste(nextmargins)), collapse=" ")
}
i <- i + 1
# Generate a row
res <- generate.row(dim=dim, subspaces=subspaces_state,
margins=margins_state, prop=prop, proptype=proptype,
dependency=dependency, discretize=discretize)
data <- rbind(data, t(res$data))
labels <- c(labels, res$label)
}
}
# Put adequate names on the columns
attributes(data)$names <- c(c(1:dim),"class")
res <- list("data"=data,"labels"=labels, "n"=n, "prop"=prop,
"proptype"=proptype, "allowOverlap" = allowOverlap,
"stream.config"=stream.config)
attr(res, "class") <- "stream"
return(res)
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.