Explainable Machine Learning Framework for Predicting Caregiver Employee Productivity and Well-Being

Abstract

Paid employment and caring for a variety of paid and non-paid care, more or less regularly, has an impact on employee well-being, engagement and productivity. This study introduces a framework for the Organizational Decision System based on the machine learning concept, to aid the support of human resources for caregivers. The proposed project differs from traditional management assessments by defining stress, flexibility in the workplace, support from supervisors, disengagement indicators, and positive indicators of enthusiasm as indicators that predict productivity indicators. This analysis employs a robust and easy-to-interpret main predictive algorithm: Random Forest classifier, suitable for the variety of employee data. Kaggle's Work from Home Employee Burnout dataset is the most useful benchmark for dataset-driven implementation, as it provides caregiver-sensitive workforce modeling relevant behavioural, workload, and burnout related variables. In order to improve productivity predictions, model features related to schedule control, emotional strain, work-care conflict and access to support were engineered. The findings of the experiment indicated that the model achieved an accuracy rate of 95.2% in predicting the category of the employees' productivity risk and well-being status. This work contributes to the unification of caregiver-responsive HR constructs and explainable machine learning for early-risk detection, intervention design and data-informed policy optimization. Thus, the suggested system provides a technically robust and performance-scalable solution for intelligent caregiver workforce management.