Penerapan K-Means Klustering dengan Algoritma Kluster Dinamik untuk Meningkatkan Kualitas Kluster pada Segmentasi Dokter Potensial: Studi Kasus PT. XYZ

The traditional K-Means algorithm has a significant weakness in its reliance on random cluster center initialization, which often results in unstable and suboptimal segmentation. This study aims to improve the quality of potential physician segmentation by proposing the integration of a dynamic clustering algorithm into the K-Means framework. The applied method is K-Means augmented with a dynamic clustering algorithm, using RFM (Recency, Frequency, Monetary) attributes derived from physician profile data. Cluster quality is evaluated using the Davies- Bouldin Index (DBI) and Purity. The proposed approach successfully improved segmentation accuracy by 25.15% compared to traditional K-Means. Quantitative analysis shows a significant improvement in cluster quality, indicated by a decrease in the DBI value from 0.846 to 0.411. Furthermore, the Purity value increased from 0.5294 to 0.7647, indicating improved cluster homogeneity. These results demonstrate that the dynamic clustering algorithm effectively addresses initialization sensitivity by iteratively adjusting cluster configurations based on inter- cluster and intra-cluster similarity. The final segmentation yielded four clusters of potential physicians with distinct RFM characteristics, enabling more targeted marketing strategies. The implementation of this model provides strategic benefits for pharmaceutical companies, including the ability to allocate promotional and sponsorship resources more efficiently based on a more accurate and reliable mapping of potential physicians.