Dhiyan Septa Wihara, Hujjatullah Fazlurrahman, Dwi Yoga Wicaksana, Achmad Kautsar, Hafid Kholidi Hadi, Muhammad Fajar Wahyudi Rahman, Dana Azizah Rahmat, Danang Ary Dewangga
Employee turnover is a persistent operational risk that undermines continuity and productivity. While many previous studies have examined turnover using traditional statistical methods, there remains a gap in linking exploratory visualization with predictive machine learning to provide actionable insights. This study couples exploratory data analysis with supervised learning to predict attrition and surface its principal drivers. The dataset contains 9,500 records from the HR department of XYZ Company, representing multiple departments and employee categories, with features including department, tenure, satisfaction, average monthly hours, salary tier, bonus, projects, performance review, and promotion. We train a Random Forest classifier and explicitly address class imbalance with Synthetic Minority Over-sampling Technique (SMOTE). On a held-out test set, the RF-SMOTE model attains strong performance (accuracy = 0.80; macro-F1 = 0.77), improving minority-class sensitivity (recall for 'Left' from 0.61 to 0.68) with a modest precision trade-off. Feature importance analyses consistently identify satisfaction, workload (average monthly hours), and tenure as the most influential predictors, while compensation variables (salary tier, bonus) and project load contribute comparatively little. The novelty of this research lies in integrating visual exploratory analysis with Random Forest modeling to uncover both descriptive patterns and predictive drivers of attrition, thereby extending existing HR analytics approaches. These findings motivate actionable interventions regular pulse surveys to monitor engagement, structured onboarding and mentoring in early tenure, and workload balancing in high-risk units to reduce voluntary separations and inform data-driven retention policy. © 2025 IEEE.
Universitas Negeri Surabaya, Faculty of Economics and Business, Digital Business Department, Surabaya, Indonesia
