Performance Evaluation of Machine Learning Algorithms for Determining Influential Factors of Antenatal Care Visits in Bangladesh

Abstract

Abstract Antenatal care (ANC) is a crucial part of the ongoing care that a mother receives before and throughout her pregnancy, at the time of delivery, and during the recovery period. This study aimed to explore the influential factors of ANC visits and evaluate the predictive model performance of identifying the determinants of ANC visits in Bangladesh using seven machine learning algorithms. This study is based on the secondary data extracted from the 2017-2018 Bangladesh Demographic and Health Survey (BDHS), which covered a nationally representative sample of 20,250 ever-married women aged 15–49 years. The final data consist of 4,946 mothers who gave birth in the three years preceding the survey. Descriptive and inferential statistical techniques along with machine learning algorithms were used for data analysis. Of the 4,946 women, most were middle-aged and in the age groups of 20-24 years (35.4%) and 25-29 years (26.2%). Receiving a greater number of ANC services was significantly positively correlated with the frequency of ANC visits. Higher wealth indices increase the chance of completing an ANC visit. The random forest (RF) model shows that age, richest, number of children, household size, and mother's primary education level are the top five important predictors of antenatal care (ANC) visits. The quality of ANC services in Bangladesh could be increased by having a better grasp of the identified risk factors and implementing them in short- and long-term initiatives.