Skip to main content

Table 1 Features selected by sequential-backward feature selection and domain experts

From: Predicting perinatal mortality based on maternal health status and health insurance service using homogeneous ensemble machine learning methods

No

Chi_best

Mi_best

SFFS

SBFS

Features Selected by Feature Selection Methods

1

Maternal age

Maternal age

ever-married sample

Community health insurance

2

Literacy

Preterm

currently pregnant

Smokes cigarettes

3

frequency of reading newspapers or magazine

highest educational level

current contraceptive method

Region

4

Family size

Religion

current use by method type

Type of cooking fuel

5

Wealth index

Ethnicity

heard family planning on radio last few months

Wealth index

6

Preterm

Family size

heard family planning on TV last few months

Current contraceptive method

7

current contraceptive method

frequency of listening to the radio

heard family planning in newspaper/magazine last few months

Hemoglobin level

8

current use by method type

type of cooking fuel

visited health facility last 12 months

Occupation

9

visited health facility last 12 months

wanted last children

Smokes pipe full of tobacco (women)

Maternal age

10

currently breastfeeding

currently breastfeeding

Chews tobacco

Marital status

11

when children put to the breast

Hemoglobin level

smokes other

Anemia level

12

Chews tobacco

anemia level

Community health insurance

Chews tobacco

13

smokes other nicotine

Marital status

Health insurance type: provided by the employer (women)

Place of residence

Accuracy

84.43

85.32

85.5

90.5

Features Selected by Domain Experts

 

14

   

Preterm

15

   

Highest educational level

16

   

Visited health facility last 12 weeks

17

   

Birth interval