Analisis Radius Pada Algoritma Birch Berdampak Terhadap Distribusi dan Kualitas Cluster

Abstract

The BIRCH method is efficient in handling large data. However, the determination of the Radius (R) parameter, which is useful for determining the maximum radius of the cluster, must be considered. R values that are too small produce many small clusters (overclustering), while values that are too large produce clusters with high heterogeneity (underclustering). The R parameter affects the distribution results and the quality of the resulting clusters can be a major problem. The lack of clear guidance in determining the optimal R value can lead to the formation of inappropriate clusters or loss of important information in the data. This study aims to provide the impact of the R value on the distribution and quality of clusters in the BIRCH. Testing several R values on employee datasets that include non-linear distribution data. Analysis is carried out to identify the relationship between the R value and the resulting data distribution pattern. The eval__uation results show that R values that are too small tend to produce over-clustering, while values that are too large cause under-clustering. The best cluster quality is achieved at a balanced R value, which is adjusted to the distribution and density of the employee dataset. Thus, it is important to choose the right R value to improve the performance of the BIRCH and ensure a representative cluster distribution. This finding provides practical guidance for adjusting the R parameter in clustering applications that implement the BIRCH.