© . .
1 "& % [7]. < ! ( " ( " ) # ! ! – " '&
" # # " # $ "/ (! # ( $) )
$) ! , # : # ! $), ), Knowledge
" # $ # # .!.,
)! !$ ! " # #
" # " , " # # # ), )! ) # "
" ), ( '& $ $ [1, 3, 7, 8]. $ # !
& ! ! ) , ) " # '& $ 00 , », '&
. ! (
$ " : # " #
) ' !. .). $ #! ! ( 0 ! # "/ #
# ! . ? ) % ( $ Data Mining ( Machine Learning ( ( " Discovery in Databases ( " ) ! $), Pattern Recognition " # ), Knowledge Extraction ( # Data Archaeology ( $ !
# $ $ # "
" ! ( !
% $ ! ! $ ,
" # ,
#! $ # # " " # #
XVII DAMDID/RCDL’2015 « 2015 .
. " !! ) 1 15 -07-03193. . . ; &
% $ ! "/ = ! # ! ! )! ! ) # . , $ # ) ). = ! & "/ ! ( !
% ; "/
# " '& ? ! #" # , " @ ! " '& ! ) " )!
"/ # '& 2.1. # # = )! #
, – ) T ( )
$ ( )
$ = ! ) " " ) )! yi .
, )! ( '& ) % # " ( # ; %
# $ x – ,
$ X –
! # ' " )
Lm :T $ x ( ).
< ! # K X y )! "/ y ,
x . " # . < ! # # ( m ! " '& $ " ) $ ! )
$ " ) – $ ! ) ! # ) X .
, ! ) " " , " # !
, '
"/ ! "/
" '& " # ! .
#" ' '& ) – , # $ ) $ "/ ,
& m .
" # ) "
c . " # !
'
P(Lm(t)) ! # !
Lm , " ! Lm
) ! «= ( Lm !
! ! # « " », . . ! " '& " # ! t1,...,tv , ».
Lm
Lm1,..., Lmz , " '& $ " )
# « $ ! " $ t1,...,tv . !! ) # " $ ! '& ! $ !
. ! $ » [14].
$ ) t1,...,tv ( 2.2. #
= ( , ( #
Y – "
) = ). = ! m ,
( "/ # '
%
X ( ( " # ' ) " ) (x, x ) )! "/
# ) X , " ! '& xa, xb yi , xc y j (*) (xa, xb)
(xa, xc) " ) t T % " # )! ? !
! "/ ! "/ # ! , % #
! '& ! # " ! = yi , y j m # & # ! " # , # ( % " # = , ( '& ) )! q
X ) " #
( ). " ! ! $ #
! ( "/ , ! # q ( '& # )! m(x, y, q) y (
' # ) $ Rmq " # yi .
" t T q* . " # q .
, $ Lm(t) q* (! ! " ) " = ,
#
PLm ( ) – $ & ! # $ #
# $ #
(
' ) ! " '& $ " " , ! # ! b {b1,...,bu} " # ! ! # ! " # " '& E(t) (!
! ' " " '& $ "
PCbb ( ) – #
- ! # = b
!
PCMm ( ) & " " [11]
# ! PCbb ( ) , # , ! , = = " = m – ' $ , , ) !
= b & PLm ( )
, . .
" ! (
% m . ;
# ! # '& ! $ " # # , ! ) ( ! ! # , # .!.) " ! !
, ) = !
' ( ( '& # ! $ !
& = , ! ! ! ( " # #
), $ $ . .
$ , " $ % ! # , " $ ( " = ! ! '& $ '& $ # ( # ! " '
! = , ! ! « #
# ' ; )! # " # " !
. " '& $ # # " » – " $ # # ( ) %
) "
T
! " #
$ " # $ ! # )! "/
xi " '& , E # ! #,
# " ( # y j ). E # " '! ' )! " )! ! ) , . .
! !
# % # % . "/ # # ( # ) $ # [11] #! m , $ «# " « « % $ ! ! " . », « ! » (« " » « ) », « # )
», ' : « ! !
# ».
# # ! = ,
( $
'& # , $ ! &
, !
# " '&
! # " ' "& $ ! #
$ ! # ) # ! ! )! ( . # ! % $ ( )! # ! # , "/ " '&
!
$ '& $
# , # , $ (# " % # " ).
( # ! (# " % # "
( ' $ # ! : " "& " ; ' ; $ .
!
! , # ! "
, # . ! " )
( # !
# ! ( # ) #
# ) ! ! '& $ !
# $ $ E " " " # " ) $N N N N
N
!
PN C $ N " $N #
N # PN # " N
N N N " ! " ! ! " ! ! ! !
N W
$N # '& 'N
ON
N N " #N ! #
N " #N
ON N N
$ #" , ! # E, $ !& $ " ! $ # ! # , $ # ! ! # . (=
! – 1: – 3: , # ! $ # )» " '! : 1 , # , » )» '& $ . ; )!
, # # '
# " , E,
# « ! ! ! ) . 1
5 –
1: ' 5
E, h= # 7-15 !! ) , # ' – ,
! #
, ! , '& ) 1 1 ! '& $ "/ " # ! "& '& !
,
$ $ 1: 2: (L , $ $ $ ! » ! " ! ( " 12 '& ' % – ! ! $ # " " % '& ( , ! # ! " !
! " # , #
$ '& $ )!
# )
$ " # ). 0 $ " # # ,
InForMedKB, $ ! ! [1] MachineLearning.ru =
& ( " ',
: < . / - , 2006. 293 .
. ., ? . ., h
$, ! , 2007. 12. . 49 -51. . . < ! ( ? " 2012., <.: « ! . .
" .
, 2014. 228 . [8] @
.s. ; " : ! «sC », 2012. 203 . [9] ; & . . @ ! # $ " :
$ $ &
! : #! !
$ // 2. <.:
. ., - $ ?
. ., ? [10] ; &
18. [11] ; & ! $ [12] ; & . ., h 0.<. < ! - $ . ? , 2003. 1 8. . 8 -18.
.L. " & $ ( . 2 h. //
. 2. <.: , 2000: 1 7. . 18 -28, 1 8. . 8
// . 2. <.: , 2012. 11. . 9 -21.
$ <.t., < ! " <.t.
:
$.
$ ! !
! ( : $ . . =! $ # ! [http://poligon.machinelearning.ru/] " i . J $ $
Sergey V. Smagin
The paper provides an introduction into the area of inductive formation of knowledge bases and highlights the current topical problems, including the interpretability of the results. A method is suggested for inductive formation of easily interpretable medical diagnostic knowledge bases. It includes the new definitions of classification and clustering problems for dependence model with parameters, the learning algorithm (solving mentioned problems in their new definitions) is developed for the practically useful and easily interpretable mathematical dependence model with parameters, which is a near real-life ontology of medical diagnostics (defined by a system of logical relationships with parameters). Also included is the software package InForMedKB (INductive FORmation of MEDical Knowledge Bases), which implements this mentioned learning algorithm.