MLP-based explainable AI model for nurses’ career fulfilment prediction

Abstract

Nurses’ career fulfilment in resource-constrained primary healthcare (PHC) systems of Western Africa significantly impacts workforce stability and healthcare delivery quality. This study introduces a novel, data-driven predictive modelling framework using a Multi-Layer Perceptron (MLP) neural network, augmented with interpretability and explainable artificial intelligence (XAI) techniques, such as quantile-quantile (Q-Q) plot, heatmap, feature importance, Williams plot, Shapley additive explanations (SHAP), and Local interpretable model-agnostic explanations (LIME), to accurately predict and interpret career fulfilment among PHC nurses from six Western African countries (Nigeria, Ghana, Senegal, Burkina Faso, Côte d’Ivoire, and Sierra Leone). A comprehensive dataset comprising 5120 PHC nurses’ responses was collected through validated instruments aligned with Herzberg’s Two-Factor Theory. Rigorous data preprocessing included normality checks, multicollinearity assessment, and robust statistical validations. The optimised MLP model demonstrated exceptional predictive accuracy with a Mean Squared Error (MSE) of 0.015, Root Mean Squared Error (RMSE) of 0.122, Mean Absolute Error (MAE) of 0.035, and R2 score of 0.970, validated through 5-fold cross-validation. XAI methods, including SHAP and LIME analyses, provided granular insights into country-specific determinants of career fulfilment, highlighting the dual role of hygiene factors and motivators, especially workplace safety, autonomy, professional development, and performance-based promotions. This study’s methodological innovations and practical insights offer targeted, actionable strategies to enhance nursing career satisfaction, retention, and healthcare quality in resource-limited settings.