contraceptive method choice

Contraceptive Method Choice

指導教授黃三益博士組員 :B924020007 王俐文 B924020009 謝孟凌 B924020014 陳怡珺

Background and Motivation

Population of the world increases tremendously, people of present day pay more attention to contraceptive method.

Step one: Translate the Business Problem into a Data Mining Problem

Topic: Contraceptive Method Choice Predict the current contraceptive method choice (no

use, long-term methods, or short-term methods) of a woman based on her demographic and socio-economic characteristics.

Especially what kind of couples would chose long-term method.

Step two: Select Appropriate Data

Title: Contraceptive Method Choice Sources:

Origin: Subset of the 1987 National Indonesia Contraceptive Prevalence Survey

Creator: Tjen-Sien Lim Date: June 7, 1997


Number of Instances: 1473

There is no missing value in this dataset.


Number of attributes:

10 (including the class attribute) Wife's age

Wife's education

Husband's education

Number of children ever born

Wife's religion

Wife's now working?

Husband's occupation

Standard-of-living index

Media exposure

Contraceptive method used (class attribute)

Step three: Get to Know the Data

Attribute Information

Attribute Name Attribute Type Description

of Attribute Value

Contraceptive method used

class attribute 1=No-use

2=Long-term

3=Short-term




of Attribute Value

Wife's age Numerical

50.040.030.020.010.0

wife_age

80

60

40

20

0

Frequency

Mean = 32.538Std. Dev. = 8.2272N = 1,473

Histogram




of Attribute Value

Wife's education Categorical 1=low 2, 3, 4=high

543210

wife_educarion

600

500

400

300

200

100

0

Frequency

Mean = 2.96Std. Dev. = 1.015N = 1,473

Histogram




of Attribute Value

Husband's education Categorical 1=low 2, 3, 4=high

543210

husband_education

1,000

800

600

400

200

0

Frequency

Mean = 3.43Std. Dev. = 0.816N = 1,473

Histogram




of Attribute Value

Number of children ever born

Numerical

15.010.05.00.0

number_of_children

300

250

200

150

100

50

0

Frequency

Mean = 3.261Std. Dev. = 2.3585N = 1,473

Histogram




of Attribute Value

Wife's religion Binary 0=Non-Islam

1=Islam

1.510.50-0.5

wife_religion

2,000

1,500

1,000

500

0

Frequency

Mean = 0.85Std. Dev. = 0.357N = 1,473

Histogram




of Attribute Value

Wife's now working?

Binary 0=Yes

1=No

1.510.50-0.5

wife_now_work

1,400

1,200

1,000

800

600

400

200

0

Frequency

Mean = 0.75Std. Dev. = 0.433N = 1,473

Histogram




of Attribute Value

Husband's occupation Categorical 1, 2, 3, 4

543210

husband_occupation

700

600

500

400

300

200

100

0

Frequency

Mean = 2.14Std. Dev. = 0.865N = 1,473

Histogram




of Attribute Value

Standard-of-living index Categorical 1=low 2, 3, 4=high

543210

standard_of_living_index

700

600

500

400

300

200

100

0

Frequency

Mean = 3.13Std. Dev. = 0.976N = 1,473

Histogram




of Attribute Value

Media exposure Binary 0=Good 1=Not good

1.510.50-0.5

media_exposure

2,500

2,000

1,500

1,000

500

0

Frequency

Mean = 0.07Std. Dev. = 0.262N = 1,473

Histogram

Step Four : Create a Model Set

Raw Data

Step Four : Create a Model Set

Total 1473 samples 75% of the data as training set

the rest of the data as testing set

→By random sampling Rapid Miner

Step Five: Fix Problems with the Data

No missing value Skewed distributions

Step Six : Transform Data to Bring Information to the Surface most of the values of the attribute named Media

Exposure are “Good” the numeric variables to do the statistical analysis to

finding outliers

1.510.50-0.5

media_exposure

2,500

2,000

1,500

1,000

500

0

Frequency

Mean = 0.07Std. Dev. = 0.262N = 1,473

Histogram

15.010.05.00.0

number_of_children

300

250

200

150

100

50

0

Frequency

Mean = 3.261Std. Dev. = 2.3585N = 1,473

Histogram

Step7 Build Model By RapidMiner, build it with Decision Tree

Step7 Build Model(con’t)

Ripper Rule if wife_age > 30 and Num_children_born <= 1 then 1 (5

3 / 1 / 3) if Num_children_born <= 0 then 1 (36 / 0 / 0) if Wife_education = 4 and wife_age <= 42 and Wife_r

eligion = 0 and Num_children_born > 3 then 2 (0 / 14 / 0)

if Wife_education = 1 and Husband_occupation = 2 then 1 (17 / 0 / 1)

if Wife_education = 4 and wife_age > 33 and Num_children_born > 2 and Husband_occupation = 1 and Num_children_born <= 3 then 2 (1 / 10 / 2)


if Num_children_born > 2 and wife_age <= 33 and Wife_now_working = 1 and Num_children_born <= 4 and wife_age > 28 then 3 (1 / 0 / 13)

if wife_age <= 35 and Num_children_born > 4 and Media_exposure = 0 then 3 (1 / 2 / 12)

if Husband_education = 4 and wife_age <= 44 and living_level = 3 and wife_age > 37 then 2 (0 / 5 / 0)

else 1 (305 / 168 / 281)


Weka-JRip (Wife_education = 4) and (Num_children_born >= 3) and (wif

e_age >= 35) => method_used=2 (178.0/76.0) (wife_age <= 33) and (Num_children_born >= 3) => method_use

d=3 (271.0/120.0) (wife_age <= 33) and (Num_children_born >= 1) and (wife_age <

= 22) => method_used=3 (106.0/51.0) => method_used=1 (771.0/342.0)


Step 8 Assess Model Decision Tree

Step 8 Assess Model(con’t) Ripper Rule

Step 8 Assess Model(con’t) JRip Rule

Conclusion

Result The problems we should improve

more data ignore some attributes details of the attribute are not so clear period and environment have changed

Thanks for you listening…

contraceptive method choice

Documents