Dynamic Clustering Analysis Using K-Means With Adaptive Centroid Optimization

Abstrak

Clustering is a fundamental technique in data analysis that aims to group data based on similarity characteristics. The K-Means algorithm is widely used due to its simplicity and computational efficiency; however, it suffers from limitations related to static centroid initialization, which may lead to suboptimal local convergence. This study proposes a dynamic clustering approach based on adaptive centroid optimization to address these limitations by allowing centroid positions to evolve iteratively according to the data distribution. The proposed method incorporates an adaptive centroid update mechanism that considers inter-cluster distances and data density, enabling a more flexible and responsive clustering process. The evaluation is conducted using multiple datasets and assessed through performance metrics such as the Silhouette Score and Davies-Bouldin Index. The results demonstrate that the proposed approach improves cluster stability and reduces misclassification compared to the conventional K-Means algorithm. Furthermore, it shows better performance when handling datasets with uneven distributions. Therefore, adaptive centroid optimization offers a promising alternative for enhancing clustering quality in various data analysis applications.