Impact of climatic factors on fertility behaviour: a machine learning approach

This study investigates the impact of climatic factors temperature, rainfall, and humidity on fertility behaviour across different socio-economic groups in Sokoto State, Nigeria, using Artificial Neural Network (ANN) and Random Forest (RF) machine learning models. Monthly time-series data from 2015 to 2024 covering climatic variables and fertility rates by rural/urban residence and employment status were analyzed. Descriptive statistics, correlation analysis, and time-series decomposition were applied. Predictive models were developed using ANN and RF algorithms, and performance was evaluated using Mean Squared Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the Coefficient of Determination (R²). Feature importance analysis identified key climatic predictors. Result shows high variability in rainfall and humidity. Correlation analysis revealed weak positive relationships between climatic variables and fertility, with humidity showing the strongest association. Both models performed poorly, with R² values between 0.0021 to 0.0650, though Random Forest (Individual) model slightly outperformed the ANN model. Feature importance analysis consistently identified humidity as the most influential factor across all population groups (Rural, Urban, Employed, Unemployed). Forecasts for 2023-2024 indicated continued disparities, with rural fertility remaining the higher and more variable than urban fertility. Climatic factors, particularly humidity, have a weak but measurable influence on fertility behaviour in Sokoto State. These findings highlight the importance of integrating climatic considerations into public health and demographic planning to mitigate the effects of environmental variability on reproductive outcomes.