International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
given continuous attribute data into discrete values. and this
operation may be a main origins of uncertainties in the whole
process of spatial knowledge discovery. At this phase, a lot of
uncertainties may be eliminated by uncertainty handling
techniques but never completely, even some new uncertainties
will be produced in handling process due to impropriety of the
techniques.
To
tT : Ll.
EEE PP < Uncertainty Origins ---
S i
M
~~
M.
u^
Uncertainties from data mining mainly refer to the limitation of
mathematical models, and mining algorithm may further
propagate, enlarge the uncertainty during the mining process.
Spatial knowledge representation exists in uncertainties,
including randomness, fuzziness and incompleteness. To a same
knowledge, it may be represented by different methods. Most of
spatial knowledge discovered by spatial data mining is
qualitative knowledge and the best way to represent them is the
natural language.
U PH...
: :
a '
’ '
’ *
1 '
' 1
5 '
’ 1
. '
* '
9 '
9 a
1 s
1 ’
; Y
; NV
: Pattern „Re
: A
: mmm Ruler ANE PA
: Fe ae AP
i a
a
| i
pas | | : :
AM | | : :
SES Lj] : :
Rnd 1
Genera- À : :
lization : :
1 à? :
Data : : :
1 2 s
4 a t
1 . .
1 1 *
1 a .
1 * 1
1 a '
1 1 .
’ 1 ’ .
a 1 4
’ 1 a 1
. ’ '
1 ' * .
1 . a
: 4 . s
1 1 i
1 1 s 1
: 1 ’ ;
’
3 H 2 1
: 4 . :
,
1 J t r
3 1 $ .
: : : :
a n 1 ’ a
HS A : :
~~" Uncertainty ^... : > i
7. Propagation —
—
-
Figure 2. Uncertainties and its propagation in the process of spatial knowledge discovery
4. SPATIAL KNOWLEDGE DISCOVERY BASED ON
FUZZY EVIDENCE THEORY
4.1 About the Evidence Theory
Evidence theory, namely Dempster-Shafer (D-S) theory, aims
to provide a theory of partial belief, which extend traditional
probability theory. Firstly, we should briefly introduce the
evidence theory.
The frame of discernment, © , is the set of mutually exclusive
and exhaustive propositions of interest. Defined on the set of
subsets of © is the basic probability assignment or mass
function, m, that associates with every subset of © a degree of
belief that lies in the interval [0, 1]. Mathematically, m is
defined as follows:
m:29 [0,1] (2)
such that:
m(d)=0 (3)
> m(x) il (4)
«co
A Belief function:
Bel( A) = > mB) (5)
BoA
A Plausibility function:
PI( A) = 1— Bel(+A) = > m(B) (6)
BOAD
Thus, at any given time the interval [ Be/( A), P/( A) ] defines
the uncertainty associated with A. While Bel(A) is the
definite support for A, P/(A) is the extent to which the
evidence at that present time fails to refute A.
When identify an object. all evidences associated with the
object must be combined. The Rule for the combination of
evidence (the Orthogonal Sum, © ):
se m, (A)m, CB)
4nriB
ls > om, (A)ym,(B)
Ao Bzo
(mm, Dm, )(c) = (7)
260
Internat.
Evidenc
intellige
knowlec
about ev
Fu:
Kn
4.2
Evidenc
randomr
randomr
vaguene
may take
theory. |
uncertaii
Herein,
reasonin
include
transforr
concept
knowled
structure
A fuzzy
the focu
operator
fuzzy se
fuzzy D
Thus, the
where th
function
m(F, ) :
The fuzz
R(r
then
where 7
B, e(l
output sy
D. wl
Hn
th
rc Bb
kind of r
evidence:
reasoning
Suppose
Then the
D-S belie
(1) Comy
(2) Mak
activatior