An Explainable AI Framework for Strategic Workforce Planning and Employee Attrition Prediction, Integrating Graph Neural Networks and SHAP-Based Counterfactual Explanations for Dynamic Career Path Modeling

• Billy Elly

  Lautech

  Author

Employee attrition represents a persistent challenge for organizations seeking to maintain workforce stability and institutional knowledge, with traditional predictive models offering limited actionable insight due to their "black-box" nature and inability to model dynamic career trajectories. This study addresses the critical gap between attrition prediction accuracy and practical interpretability by developing an integrated framework that combines Graph Neural Networks (GNNs) for career path modeling, SHAP-based counterfactual explanations for actionable insights, and multi-task learning for simultaneous workforce demand forecasting. Using a retrospective dataset of 1,470 employee records augmented with synthetic career trajectory data, the framework achieves 89.4% prediction accuracy with an AUC of 0.931 and F1-score of 0.887, substantially outperforming baseline models including Random Forest (86.2%) and Logistic Regression (79.1%). SHAP analysis identifies job satisfaction, overtime frequency, and career progression velocity as primary attrition drivers, while counterfactual explanations generate personalized retention recommendations with an average of 72% predicted adherence improvement. The framework's dual-output design—providing both organizational-level workforce forecasts and individual-level career path projections—enables strategic planning at multiple decision horizons. This research contributes a replicable methodology for explainable, graph-based workforce analytics and demonstrates how counterfactual reasoning can transform predictive insights into prescriptive HR interventions.

• pdf

Copyright (c) 2026 Billy Elly (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.