Analyzing and forecasting under-5 mortality trends in Bangladesh using machine learning techniques.

Area of Science:

• Public Health
• Demography
• Data Science

Background:

• Under-5 mortality is a key indicator of development, especially in Bangladesh.
• Machine learning models are utilized to forecast under-5 mortality trends.
• Actionable insights are provided for policymakers and health professionals.

Purpose of the Study:

• To forecast future trends in under-5 mortality in Bangladesh using machine learning.
• To identify the most accurate machine learning model for predicting under-5 mortality.
• To provide data-driven recommendations for public health interventions.

Main Methods:

• Analysis of Bangladesh Demographic and Health Survey (BDHS) data from 1993-94 to 2017-18.
• Application of various machine learning algorithms including Linear Regression, XGBoost, and CatBoost.
• Model performance evaluation using metrics like MAE, RMSE, R-squared, and MAPE, with k-fold cross-validation.

Main Results:

• A significant decline in under-5 mortality in Bangladesh was observed from 1994 to 2018.
• Linear Regression model demonstrated highest accuracy with lowest error metrics and highest R-squared.
• Projections indicate continued reduction, reaching 29.87 by 2030 and 26.21 by 2035 per 1,000 live births.

Conclusions:

• Under-5 mortality in Bangladesh decreased by 76.72% from 1994 to 2018.
• Linear Regression model accurately forecasts under-5 mortality trends.
• Forecasted rates may not meet the SDG target, necessitating enhanced healthcare and maternal health interventions.