Enhancing Workforce Attendance Evaluation in Vocational Schools: A Decision Support Framework Powered by Explainable Machine Learning

Abstract

This study addresses the underexplored potential of integrating explainable machine learning techniques into Decision Support Systems (DSS) for workforce attendance evaluation, specifically in vocational high schools. While prior research often emphasizes predictive modeling, this paper proposes a non-predictive, descriptive framework that enhances interpretability and actionability. Utilizing a dataset of 52,000 biometric attendance records collected from five vocational schools over six months, the study applies K-Means clustering, anomaly detection (Isolation Forest, LOF), and SHAP-based feature importance to uncover patterns, irregularities, and key drivers of attendance behavior. The findings reveal three distinct behavioral clusters, significant anomaly segments, and key influencing variables, such as workload, weekday cycles, and seasonal peaks. These insights are operationalized into an interactive DSS framework that supports real-time segmentation, alerts, and decision-making for school administrators. Unlike conventional DSS approaches, the proposed system is equipped with explainable outputs and visual dashboards, increasing transparency and decision relevance. The study’s contributions include: (1) an explainable ML-enhanced DSS architecture tailored for workforce evaluation, (2) a shift from predictive modeling to descriptive, interpretable analytics, and (3) empirical evidence linking attendance trends to workforce performance metrics. This framework offers scalable applicability for educational institutions aiming to adopt data-driven, interpretable decision support tools in human resource management.