Abstract
Objectives
This study aims to signify the best classifier to predict stunting with the comparative scenario between three South Asian countries that will help mitigate the urgency of addressing child stunting during childhood.
Methods
The DHS datasets like BDHS 2017-18, IDHS 2019-21, and NDHS 2016 had been used here to extract the necessary information for measuring child stunting. After completing inevitable parts, frequency table and chi-square had been used to present the compared scenario and the prediction of child stunting was performed with different machine learning algorithms.
Results
The prevalence of stunting is 28%, 33.1%, and 32.9% for BD, IN, and NP respectively. The result indicates that 53% stunted children are male in India (p < 0.01), but not significant in BD and NP. Moreover, 68% Nepali stunted children did not have baby postnatal checkup (p = 0.014). In addition, immunization status was only significant in Bangladesh (p < 0.01). The RF classifier outperformed among all the classifiers with 77.66%, 62.45%, and 74.81% accuracy score for BD, IN, and NP respectively.
Conclusion
The country-wise prevalence of child stunting with the associated factors is highlighted by this study. Moreover, to detect stunting early, this study suggests using the RF classifier for all the country. The findings of this study will help the policy makers and the other agencies to take the immediate step to reduce child stunting and make the world better for the next generations by the early detection of malnutrition using the classifier.