Information Technology Reference
In-Depth Information
Definition 7.18
Algorithm index table AI is a set of names of all algorithm
α
registered in BSDT. Algorithm set
Α
is corresponding to algorithm index table
AI one-to-one.
Definition 7.19
Case attribute description set
Θ
is a six-element tuple <
e
d
,S
t
a
π
,S
> defined in classical case base, where
π
denotes domain name
of classical training examples set E
,S
,
η
, N
t
e
handled by algorithm
δ
; S
represents size
t
d
t
t
of E
; S
represents dimension of E
; S
represents time of generating decision tree
t
which accords with E
through algorithm
δ
;
η
represents error-classification
ratio of decision tree; N
a
a
represents name of algorithm
δ
. Here N
is taken as a
class sign of case base.
Definition 7.20
Retrieval information tuple
θ
is a three-element tuple <a
1
,a
2
,a
3
>
defined in primitive training example set, where a
1
represents domain name of
given primitive training example set; a
2
represents size of E; a
3
represents
dimension of E.
Definition 7.21
Case base CE is a set of cases generated by BSDT algorithm. Its
six attributes are described by case attribute description set; category N
a
∈
algorithm index table AP; example ce
∈
CE is generated by algorithm ANA.
Definition 7.22
Domain name
π∈
domain name set
Π
. In BSDT,
Π
={1-agriculture, 2-industry, 3-commence, 4-education, 5-electronics, 6-physics,
7-chemistry, 8-mathematics, 9-medicine, 10-others}.
Definition 7.23
Classical case base TEB = example index table EI
∪
classical
case table TET
。
Example index table EI is a set of all domain names
π
registered in algorithm BSDT; classical example table TET is a set of examples
which get from a domain and named by a domain name
π
.
Definition 7.24
An optimal index
ζ
is a case retrieval standard decided by both
case attribute description set
Θ
and retrieval information tuple
θ
. It is
determined by following formula:
e
d
t
ζ
=
(
a
=
π
)
∧
(
a
−
S
<
λ
)
∧
(
a
−
S
<
λ
)
∧
(
S
•
η
<
λ
)
1
2
1
3
2
3
where λ
1
λ
2
λ
3
are tuple threshold, dimension threshold and tree controlling
threshold respectively, and can be tuned in running time. In BSDT, default value
