THE APPLICATION RESEARCH OF KNOWLEDGE DISCOVERY TECHNIQUES 
BASED ON ROUGH SET IN DECISION SUPPORT 
WANG Shaohua *, BIAN Fulin* 
? Research Center of Spatial Information & Digital Engineering, Wuhan University, Hubei,P. R. China, (430079) 
snoopywsh@163.com 
Commission VI, WG II/5 
KEY WORDS: GIS,Knowledge,Spatial, Rough set,Decision support 
ABSTRACT 
With the application and development of science and technology, the tremendous amount of spatial and nonspatial data have been 
stored in large spatial data bases. Analysing them for decision is badly in need of spatial data mining and knowledge discovery to 
provid knowledge. In recent years, some efforts in knowledge discovery have focused on applying the rough set method to 
knowledge discovery. In this paper, the application of knowledge discovery method based on rough set in land use decision support 
system is discussed. First the characteristic and development of knowledge discovery method based on rough set are briefly stated. 
Second, the characteristic of spatial data in GIS is discussed. Third, a knowledge discovery method based on rough set is put forward 
in land use decision support system. The procedures for this method, the algorithm and key matters are also analyzed. Finally, rules 
extracted by the method shows a good result. This method has solved the problem of obtaining the decision rule in DSS effectively. 
1. INTRODUCTION 
With the rapid development of the technology of data 
acquisition and database,extensive data in geographical 
information systems is increasing constantly. The present GIS 
systems mainly have such functions as data inputting, inquiry 
and statistics of those data,etc..Hence, their analysis functions is 
still very weak and not flexible, and cannot find relations and 
rules in the data effectively, so it is very difficult to extract the 
implicit mode to solve the problem on complicated spatial 
decision. 
Data mining technology provides a new thoughts for organizing 
and managing tremendous spatial and nonspatial data. Rough 
set theory is one of the important method for knowledge 
discovery,which was firstly put forward by Pawlak in 1982. 
This method can analyze intactly data, obtain uncertain 
knowledge and offer an effective tool by reasoning. 
GIS contains large amount of spatial and attribute data,and its 
information is more abundant and complicated than those stored 
in general relation databases and affair databases. The 
application of rough set theory to discover knowledge for 
dicision in spatial database is increasingly important in the 
construction of GIS system. 
Taking the land use dicision support system as an example,this 
paper presents a method for knowledge discovery in spatial 
databases,on the basis of rough set theory, and illustrates its 
process. 
2. OVERVIEW 
2.1 GIS and Decision support 
The geographical information system began to develop rapidly 
in 1960s. The most important characteristic of GIS technology 
is integrating and managing tremendous multi-subject spatial 
and attribute data.It can connects such attribute information as 
the society, economy, population, etc. with spatial position of 
the earth surface to establish a complete decision information 
254 
database for inquiry, analyse and display. The rapid 
development of information technology and the new 
requirments in this field has revealed the defects of affair- 
oriented GIS which is urged to tranfer from management to 
dicision support. 
An important trend of GIS development is intelligent DSS 
s(IDSS)(Li Deren,1995).GIS are increasingly being used for 
decision-making, yet it is still not enough to solve semi- or ill- 
structured decision problems that own the character of fuzziness 
and uncertainty. This makes the study on knowledge-based GIS 
interesting to researchers (Cohena & Shoshany, 2002; Y amadaa, 
et al., 2003). Knowledge is the foundmental of. DSS,and many 
researchs are being focused on discoverying knowledge from 
tremendous database. 
2.2 Spatial Data Mining 
Spatial data mining (also called geographical knowledge 
dicovery),which is a branch of data mining, puts emphasis on 
extracting implicit. knowledge,spatial relations and other 
significative modes from spatial data. 
Different from general mining tasks,spatial data mining is 
mainly involved in the researches on the probability distribution 
modes, clustering and classification characteristics,reliance 
relation between attributes etc. of spatial data. It is much 
complicated than general data mining in relation 
database,which is shown in two aspects. 
e Huge amount of spatial data and the complexity of spatial 
data type and spatial visit method. That is,besides the 
nonspatial information such as word,characters, spatial 
data,contains the spatial information such as topological 
relations, distance relation and direction relation. 
e The relations between spatial data are connatural.The 
relation between spatial entities is connatural,so is the 
relation between such attributes as population,economy and 
social development in spatial entities, which makes the 
spatial data mining more difficult. 
Rough set,consisting of upper approxmate set and lower 
approxmate set, is a tool for intelligent dicision analysis dealing 
Internal 
with in 
suitable 
attribute 
and knc 
of rou 
consiste 
of attrit 
data, at 
the rel: 
uncertal 
minimu 
3.1 Ba 
Assume 
an equi 
categori 
not be « 
subsets : 
can be 
contrary 
defined 
Routh s 
set: 
Lower / 
R (X): 
Upper A 
R(X)- 
Boundai 
bn ,(X 
posR(X 
while ne 
3.2 Kn 
Knowle 
expressi 
knowled 
themseh 
À know] 
where U 
CUT 
attribute 
constraii 
attribute 
Vis the 
is the fic 
1x4 
Let attril 
B 
Define ( 
U— Vy ti 
R upper 
expresse 
R upper 
expresse 
Define { 
transacti 
Then the 
transacti