THE USE OF CLUSTERING MACHINE LEARNING METHOD IN DETERMINING MEDICAL CHECK UP (MCU) HEALTH PARTICIPANTS AT PT PLN (PERSERO) KALIMANTAN REGION

• Main Author: Nikma Srihandayani, Andi
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/86704
• Institution: Institut Teknologi Bandung

Health is a crucial aspect in supporting work productivity, as optimal performance can only be achieved when employees are in good health. PT PLN (Persero) is one of the companies that places significant concern on the health of its employees, ensuring that they can work under the best possible conditions. The provision of health facilities for employees, retirees, and their families is a testament to the company's commitment to the well-being of its human resources. These health facilities include preventive, curative, and rehabilitative measures. One of the preventive actions implemented by the company is the annual provision of medical check-up (MCU) facilities for employees. However, the implementation of the MCU program is currently limited to specific job categories and age groups. This highlights the need for a more systematic strategy to prioritize MCU beneficiaries based on health and employment data. In line with this, PT PLN (Persero) maintains a database containing various important employment-related information, including employee personal data, career history, attendance records, and health records. However, the utilization of this database to support employee health management has not yet been effectively optimized. Systematic and integrated management of employee data has the potential to produce valuable strategic information. This information is not only useful for gaining a deeper understanding of employee health conditions but can also serve as a basis for more accurate decision-making in designing and developing future health programs for employees. Optimizing the use of this database is expected to improve the efficiency and effectiveness of sustainable health management. The use of machine learning methods, particularly clustering, is necessary to process data with the aim of determining employees eligible for medical check-up (MCU) benefits based on employment and health data. This study aims to recommend the most optimal clustering method for grouping MCU participants among employees in the Kalimantan Regional Office. The data utilized include employee personal information, job history, attendance records, and health data, which consist of health transaction records and MCU history. Optimizing data management is expected to enhance both the efficiency and accuracy of decisionmaking processes related to MCU beneficiary selection. The analysis is carried out by implementing various clustering methods, including K-means, K-medoids, Gaussian Mixture Model (GMM), Agglomerative Hierarchical Clustering, and DBSCAN. Furthermore, each clustering method is applied with dimensionality reduction using Principal Component Analysis (PCA) to improve the efficiency of the clustering process. This study also leverages Variational Autoencoder (VAE) to produce better latent representations, aiming to obtain optimal and relevant clustering results for the research objectives. The evaluation of cluster quality is conducted using the Silhouette Score metric to determine the clustering method that delivers the best results. This assessment ensures that the resulting cluster structure has high validity and interpretability. The research results indicate that the implementation of the K-medoids clustering method, combined with Principal Component Analysis (PCA) and Variational Autoencoder (VAE), produces the most optimal outcomes across all cluster tests conducted, including groupings into two, three, four, and five clusters. In the twocluster scenario, this method achieves a Silhouette Score of 0.805, indicating a strong and superior cluster structure compared to other clustering methods. Based on these findings, it can be concluded that the K-medoids clustering method, combined with PCA and VAE, is the most effective approach for grouping MCU data. This approach provides more accurate recommendations regarding employees eligible for medical check-up (MCU) benefits.