Forecasting Annual Maternal Mortality Rate For Kazakhstan Using The Multilayer Perceptron

Abstract

Maternal mortality remains a pressing global public health challenge, with nearly 800 women dying daily from preventable pregnancy-related causes, predominantly in low- and lower-middle-income countries. While global maternal mortality ratios (MMR) declined by 34.3% between 2000 and 2020, significant disparities persist, with low-income nations experiencing MMRs over 40 times higher than those in Europe. Kazakhstan, despite notable socioeconomic progress and targeted health interventions, continues to underperform relative to OECD counterparts in maternal health outcomes, prompting the need for evidence-based forecasting to guide policy responses. This study employs annual time series data on Kazakhstan’s MMR from 1985 to 2023 to model and predict future trends for the period 2024 to 2035. Using a Multilayer Perceptron artificial neural network (ANN) approach, specifically the ANN (12, 12, 1) architecture with hyperbolic tangent activation function; the research capitalizes on the model’s flexibility and capacity for nonlinear pattern recognition. Data were sourced from the World Bank online database, and model stability was rigorously evaluated using residual analysis and forecast evaluation criteria, including Mean Squared Error (MSE) and Mean Absolute Error (MAE). The findings show that the applied neural network model exhibits strong stability, with an MSE of 2.643 and MAE of 1.029, supported by favourable residual diagnostics. Out-of-sample projections indicate that Kazakhstan’s annual MMR will remain consistently low throughout the forecast horizon, with predicted values ranging from 10.38 to 14.04 per 100,000 live births between 2024 and 2035. These projections suggest a sustained decline from historical peaks, including the spike observed in 2020 (36.5 per 100,000 live births). The study points to the importance of continued investment in maternal health interventions, targeted efforts to address socioeconomic determinants such as secondary education enrollment and domestic violence, and strengthening healthcare systems to maintain this positive trajectory. Policymakers and health authorities in Kazakhstan are encouraged to persistently address local drivers of maternal mortality while leveraging predictive analytics for proactive resource allocation and strategic planning.