Exploring the achievements and forecasting of SDG 3 using machine learning algorithms: Bangladesh perspective

• Published in: PloS one Vol. 20; no. 3; p. e0314466
• Main Authors: Molla, Md. Maeen, Hossain, Md. Sifat, Ali, Md. Ayub, Islam, Md. Raqibul, Sultana, Mst. Papia, Roy, Dulal Chandra
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 04.03.2025; Public Library of Science (PLoS)
• Subjects: Algorithms; Analysis; Autoregressive models; Bangladesh - epidemiology; Biology and Life Sciences; Births; Child Mortality - trends; Child, Preschool; Data mining; Death; Ecology and Environmental Sciences; Fatalities; Female; Forecasting; Forecasts and trends; Health aspects; Humans; Indicators; Infant; Infant mortality; Infant Mortality - trends; Infant, Newborn; Infants; Infants (Newborn); Learning algorithms; Machine Learning; Management; Maternal mortality; Maternal Mortality - trends; Medicine and Health Sciences; Methods; Mortality; Mothers; Neonates; Neural networks; Neural Networks, Computer; NMR; Nuclear magnetic resonance; Patient outcomes; People and Places; Physical Sciences; Recurrent neural networks; Research and Analysis Methods; Root-mean-square errors; Sustainable development; Sustainable Development - trends; Technology application; Time series; Bangladesh
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0314466

Sustainable Development Goal 3 (SDG 3), focusing on ensuring healthy lives and well-being for all, holds global significance and is particularly vital for Bangladesh. Neonatal Mortality Rate (NMR), Under-5 Mortality Rate (U5MR), Maternal Mortality Ratio (MMR) and Death Rate Due to Road Traffic Injuries (RTI) are considered responsible indicators of SDG 3 progress in Bangladesh. The objective of the study is to forecast these indicators of Bangladesh up to 2030 and compare these forecasts with predetermined 2030 targets. The data is obtained from the World Bank's (WB) website. For forecasting, time series models were employed, specifically Autoregressive Integrated Moving Average- ARIMA (0,2,1) with Akaike Information Criterion (AIC) 94.6 for NMR and ARIMA (2,1,2) with AIC 423.2 for U5MR, selected based on their lowest AIC values. Additionally, Machine Learning (ML) models, including Bidirectional Recurrent Neural Networks (BRNN) and Elastic Neural Networks (ENET), were employed for all the indicators. ENET demonstrates superior performance compared to both BRNN and ARIMA in the context of NMR, achieving a Root Mean Absolute Error (RMAE) of 0.603446 and a Root Mean Square Error (RMSE) of 0.451162. Furthermore, when considering U5MR, MMR, and Death Rate Due to RTI, ENET consistently exhibits lower error metrics compared to the alternative models. Following the time series and ML analyses, a consistent trend emerges in the forecasted values for NMR and U5MR, which consistently fall below their respective 2030 targets. This promising finding suggests that Bangladesh is making significant progress toward meeting its 2030 targets for NMR and U5MR. However, in the cases of MMR and Death Rate Due to RTI, the forecasted values exceeded 2030 targets. This indicates that Bangladesh faces challenges in meeting the 2030 targets for MMR and Death Rate Due to RTI. The analyses underscore the importance of SDG 3 in Bangladesh and its progress towards ensuring healthy lives and well-being for all. While there is optimism regarding NMR and U5MR, more focused efforts may be needed to address the challenges posed by MMR and Death Rate Due to RTI to align with the 2030 targets. This study contributes valuable insights into Bangladesh's journey toward sustainable development in the realm of health and well-being.