Skill mismatch remains a persistent challenge in Indonesia’s labor market, particularly in East Java, where 48.51% of workers experience mismatch. Traditional econometric methods often lack predictive accuracy and real-time applicability for policy-making. This study introduces a machine learning (ML) approach to enhance mismatch prediction using Sakernas 2024 data. By applying ensemble tree-based algorithms such as LightsGBM, Random Forest, and Decision Tree are implemented to optimize predictive performance. The best-performing model, LightGBM, achieves 98.3% accuracy, 98.1% precision, and 98.0% recall. Key predictors include job category, education attainment, age, and salary. The findings are used to propose a roadmap for developing a web-based decision support system that offers real-time guidance for seekers, educators, and policymakers. This system isenvioned to support the broader Indonesia Emas 2045 agenda by enabling evidence-based interventions, improving workforce planning, and narrowing the education-employment gap through intelligent prediction.

Keywords: Labor market, machine learning, skill mismatch.