Professional Documents
Culture Documents
5
Sinh vin: Lu Cng T Ngi hng dn: V Tin Thnh
Outline
Thut ton Cy quyt nh
nh ngha. Xy dng cy quyt nh. c im cy quyt nh.
Node L
Node L
Node L
V d cy quyt nh
4
im yu ca cy quyt nh:
D xy ra li khi c nhiu lp: do ch thao tc vi cc lp c gi tr dng nh phn. Chi ph tnh ton t hc: do phi i qua nhiu node n node l cui cng
S={S1,S2,,Sx)
10
Test B s c chn nu c G(S, B) t gi tr ln nht. Thng tin tim nng (potential information) ca bn thn mi phn hoch:
11
12
X l nhng gi tr thiu:
L case c thuc tnh khng c gi tr. information gain v potential information: S0 = case thuc S c gi tr thuc tnh = null.
13
S dng c ch lu d liu thng tr trong b nh => ng dng vi database nh ( tn s li lp li 4% vi database 20000 cases). C c ch x l thiu, li hoc qu va d liu. Lut to ra n gin.
15
Bc 2 (Phn loi): da trn m hnh xy dng phn lp d liu mi: i t gc n cc nt l nhm rt ra lp ca i tng cn xt.
16
17
C4.5 v C5.0
Lut: C5.0 chnh xc hn, nhanh hn, tn t b nh hn.
Blue: C5.0
18
C4.5 v C5.0
Boost: to v kt hp nhiu lp phn loi tng chnh xc d on
Blue: C5.0
19
Hng nghin cu
Cy n nh:
Tn s li ca cy c xy dng t data case c cu trc thp hn nhiu so vi data case khng nhn thy c. VD: vi 20k case c cu trc, t l li l 4%,
Cng 20k case v c 1 case khng c kim tra, t l li 11,7%
22