generate.static.stream: Generate Static Data Stream

Description Usage Arguments Details Value Author(s) See Also Examples

Description

Generate a new synthetic multidimensional static data stream having the desired properties.

Usage

1
2
generate.static.stream(n = 1000, prop = 0.01, proptype = "proportional",
  stream.config = NULL)

Arguments

n

A vector containing x values, where the values corresponds to the number of points for each step and x to the number of steps.

prop

Proportion of outliers in the hidden space.

proptype

Type of the proportion of outliers. Value "proportional": depend on the size of the empty space. Value "absolute": same absolute proportion per subspace.

stream.config

A stream configuration object. Should have been generated with nstep = 1.

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 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.

Value

An object of class stream, which is a List of 5 elements.

Author(s)

Edouard Fouché, edouard.fouche@kit.edu

See Also

Examples

1
2
3
4
5
6
7
# 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")

edouardfouche/R-streamgenerator documentation built on May 15, 2019, 11:02 p.m.