#relative: rule-36,rule-29,rule-17, 
rule-90,rule-103; 
/* rules connected with #rule-81 */ 
The RELATIVE of a rule is one or several 
other rules that have almost the same 
conditions, and is used to increase searching 
speed. This item may be got automatically by 
the system. 
3. REASONNING METHODS USED IN MCGES-GIE 
In accordance with the three levels of 
knowledge, a complete inference procedure of 
MCGES-GIE is composed of several operations 
in three levels. 
3.1 Basic inference 
We can use the knowledge in the first level 
to match values of factors got from MCGIS , and 
grade the factors. The results of basic 
inference are the classes of basic 
geographical factors of objects. 
3.2 Medium inference 
There are two goals in this step, one is to 
calculate attribute values of integrated 
geographical factors, another is to classify 
the integrated geographical factors based on 
the conclusions of the basic inference and 
the medium knowledge (rules). 
Now we can get a great deal of remote sensing 
data,but much less survey data.Remote sensing 
data may be considered as a kind output of 
geographical phenomena, and those relatively 
less survey data may be considered as the 
really geographical features. Fuzzy 
mathematics is a suitable tool to distinguish 
successive varibles and can be conveniently 
used to build geographical classification 
model with remote sensing data.For example,if 
we have got MSS-4,5,6 and 7 image data of the 
same district, we can construct a fuzzy 
function to classify landuse types in this 
district according to the image data of four 
bands, 
object set: X={x ! x'=(x1,x2,x3,x4)} 
where x1,x2,x3,and x4 represent grey values 
of MSS-4,5,6,and 7. 
fuzzy classification model is 
AizAilfl Ai2 f] Ai3 f] Ai4 (1) 
where Aij is normal fuzzy set, 
jurisdiction function of Ai on X is 
Ai(x)= min Aij(xj) = 
I<=j¢<=4 ^" 
xj-aij \2 
exp [: max —————— ) J (2) 
1<=j<=4 bij 
where aij=Exij, bj= dDxij; 
izl, 2,..., n are type code; 
j=1, 2, 3, 4 are band code. 
then we have following classification 
267 
formula, 
a. A (x) max min Aij (xj)- 
ek 1<=i<=4 1<=j<=4 ~ 
xij-aij \2 
bn. Je 
1<=i<=n 1<=j<=4 bij 
4 
b. (Ai,B)= A (Aij,Bj)= min (Aij,Bj) (4) 
ths j=1 " - 1<=j<=4 
if let ej=Exj, dj=ADx, xj € Bj, 
ew ^ 
4 
and B- f| Bj is the fuzzy subset that need 
be classified, then we have, 
(Ak,B)= max (Ai,B) - 
"e deje ^ 
aij-ej à 
0.5* | 1texpi- min max ME (5) 
1<=i<=4 1<=j<=4\ bij+dj 
and B can be considered as type k. 
Grey system theory can forcast the developing 
trendency of varibles with relatively less 
known condition, and can describe the 
unbalanced relationship between main varible 
and subordinated varibles. Helped by grey 
system theory, we can build geographical grey 
model based on remote sensing data and survey 
data.A general Grey Model (GM)may be shown as 
GM(n, h) , which is a n factorial, h varibles 
differential equation,its expression is, 
n,Q) h-t,, CO T 
d e + Qu M AS t+ Qa x, 
= bı y C " b, x ++ Du vm (6) 
After getting integrated geographical factor 
values,using medium knowledge and conclusions 
of basic inference to match with them, we can 
gain the geographical classification results, 
which will be used in advanced inference. 
3.3 Advanced inference 
Based on the conclusions of the basic and 
medium inference and the advanced knowledge 
(rules) ,we can gain the decision measures or 
divisions and planning scheme on a certain 
geographical problem, which is the last step 
in the inference. 
Because most of geographical inference are 
shown as under certain conditions to gain 
certain results,in this step,MCGES-GIE adopts 
following production strategy in reasonning: 
RULE : if A then B 
PREMISE : A is true 
CONCLUSION : B is true 
EFFECTIVENESS: possibility 
The advanced inference uses Bayes theorem and 
fuzzy logic to reason. If we record the 
probability of conclusion with evidence 
existing as P(CiE) ,then Bayes theorem may be 
show as,