Dicision Tree (Thuật toán phân lớp)

Ví dụ 1:  xây dựng cây quyết định từ bảng sau:

Đáp án

Ví dụ 2:

Giải

Áp dụng các công thức sau:

Entropy : E = SUM(-pi * log2(pi))

Gain : G = E - SUM(Si/S * Ei)

 

Find Root

9+, 5- E = 0.94
Age	Income	Student	Credit_rating
Youth 2+, 3- E = 0.97   Middle_aged 4+,0- E = 0   Senior 3+, 2- E = 0.97   G= 0.94 - 5/14 * 0.94 - 0 - 5/14 * 0.97 = 0.247 (max)	G = 0.029	G = 0.152	G = 0.048

=> Root is Age (Youth, Middle_aged, Senior)

 

 

 

 

Find node for brand Youth (brands: youth)

2+, 3- E = 0.97
Income	Student	Credit_rating
High 0+, 2- E = 0   Medium 1+, 1- E= 1   Low 1+, 0- E = 0   G = 0.97 - 2/5 * 1  = 0.57	G = 0.971 (max)	G = 0.02

=> Brand Youth have node is Student (no, yes)

 

Find node for brand no (brands: youth - no)

0, 3- E = 0

=> Brand no have leaf is NO

 

Find node for brand yes (brands: youth - yes)

2+, 0- E = 0

=> Brand yes have leaf is YES

 

 

Find node for brand Middle_aged (brands: middle_aged)

4+, 0- E = 0

=> Brand Middle_aged have leaf is YES

 

Find node for brand Senior (brands: senior) //i forget student, you need add student

3+, 2- E = 0.97
Income	Credit_rating
G = 0.0202	G = 0.971 (max)

=> Brand Senior have node is Credit_rating (fair, excellent)

 

Find node for brand fair (brands: senior - fair)

3+, 0- E = 0

=> Brand fair have leaf is YES

 

Find node for brand excellent (brands: senior - excellent)

0+, 2- E = 0

=> Brand excellent have leaf is NO

 

Tree:

Root			Age
Brand	Youth		Middle_aged	Senior
Node	Student			Credit_rating
Brand	Yes	No		Fair	Excellent
Leaf	YES	NO	YES	YES	NO

 

Tôi quên sinh viên, bạn cần sinh viên Caculator