DESCRIPTION

r.roughset is the implementation of the rough set theory [1,2] in GRASS GIS environment. It requires the following input:
1. the geographical attributes constituting the information system for the rough set analysis; they have to describe environmental, economic or social issues(attributes=string[,string,...]);
2.the theme in which areas with the issues to be studied are identified (decision=string).

An information system is generated and the functions of version 2 of rough set library (RSL, ver. 2.0) [3] are applied, following one of the decision strategies available with the optionstrgy=string.

The choice of the classification strategy, clssfy=string is necessary to generate a map based on the attributes and the decision rules. Three options are available:Classify1,Classify2,Classify3 each of them is referred to a different algorithm. [3].

It is possible to get three different output files; two of them are text files, the third is a raster file:

a. text file structured following the standard of the rough set library, having the name registered in the outTXT=string without any additional extension. The file can be used as well with RSL implementation out of GIS environment.

NAME: name of the information system

ATTRIBUTES: 4 (number of attributes)

OBJECTS: 8 (number of objects)

0 0 0 0

1 0 1 1

1 1 0 0

0 2 0 1

1 2 0 1

1 0 0 0

1 2 0 1

0 0 1 1

The number of attributes is the number of raster given as input at attributes=string[,string,...] plus the decision raster (decision=string) located in the last column of the information system, by default.

b. text file, having the name defined in outTXT=string and the .out, extension. It contains the decision rules extracted from the geographic information system; decision rules are expressed as:  if ... then and they are easy to read and to be interpreted by the user;
c. raster, having the name defined in the field outMAP=string, resulting from the application of the decision rules generated by the geographic information system following one of the classification strategies in clssfy=string. Rules to apply can derive from an information system generated from the data entered in attributes=string and decision=string or from a text file generated from a previous processing stage. In this case the file structure must be as the one reported in box 1.

NOTES

The module can process CELL raster, and calculations can be made only with integer (int) variables. This is a constraint imposed by the rough set library. The decision rules, therefore, provide only integer information.

REFERENCE

  1. Pawlak Z. Rough Sets, International Journal of Information and Computer Science Vol. 11, No. 5, 1982, pp. 344-356.

  2. Pawlak Z. Rough Sets, Theoretical Aspects of Reasoning about Data, Kluwer Academic Publishers, 1991.

  3. Gawrys M., Sienkiewicz J. Rough Set Library - User's manual (ver. 2.0), September 1993

  4. http://en.wikipedia.org/wiki/Rough_set - “rough set”

SEE ALSO

r.mcda.fuzzy, r.mcda.electre, r.mcda.regime

AUTHORS

Antonio Boggia - Gianluca Massei
Department of Economics and Appraisal - University of Perugia - Italy

Last changed: $Date$