J. Handl: data generators

Syntetic data generators for creating datasets with Gaussian distribution. The code was taken from official website and slightly modified for modern compilers.

Requirements

You'll need g++ compiler.

• Debian: apt-get install build-essential should be enough

and run

$ make

mult_generator

Currently does not take any parameters, all settings is hard-coded in constants:

#define DIM 2        // dimensionality of the data
#define NUM 40      // number of clusters

#define MAXMU 10    // mean in each dimension is in range [0,MAXMU]
#define MINMU -10
#define MINSIGMA 0
#define MAXSIGMA 20*sqrt(DIM) // standard deviation (to be added on top
// of row sum in each dimension is in range [0,MAXSIGMA]
#define MAXSIZE 100  // size of each cluster is in range [MINSIZE,MAXSIZE]
#define MINSIZE 10
#define RUNS 10      // number of data sets to be generated

simply run:

$ ./mult_generator

elly

Ellipsoid generator

$ ./elly [-k <nclust>] [-d <dimension>] [-s <seed>]

where all parameters are optional and:

• <nclust> is a positive int >= 2
• <dimension> is a positive int >= 2
• <seed> is a long int.

cure

CURE data sets generator. See Guha, Sudipto, Rajeev Rastogi, and Kyuseok Shim. "CURE: an efficient clustering algorithm for large databases." ACM SIGMOD Record. Vol. 27. No. 2. ACM, 1998. for more details.

The distribution of data points is just approximated

$ ./cure -n <npoints> [-d <dimension>] [-s <seed>] [-l <x/y min>] [-m <x/y max>] [-t type of data]

where:

• -l minimal x/y value
• -m maximal x/y value
• -t type of dataset, currently supports values 0-2

disk

Disk in disk dataset are two clusters formed by a circle and an annulus around it.

Authors

• Julia Handl
• Joshua Knowles
• John Burkardt
• Tomas Barton