Studi Perbandingan Kinerja Pra-proses Data dalam Pengklasteran Potensi Investasi Negara Berkembang Menggunakan Agglomerative Hierarchical Clustering

Abstract

Investors require information to consider when making decisions related to investing in developing countries. This is done to assess the associated risks as well as the potential for achieving the expected returns. Based on supporting aspects related to the level of investment, a clustering process is conducted to group developing countries so that investors can more effectively evaluate both risks and potentials. The clustering method used in this study is agglomerative hierarchical clustering. Data pre-processing plays a crucial role before the clustering process begins; therefore, a comparative study of various pre-processing methods is necessary to achieve the most optimal clustering results and to provide a foundation for further research. The findings indicate that the best clustering result was obtained through standardization pre-processing with PCA (super-cluster division using single linkage), yielding a Silhouette Score (s(x_i)) of 0.6503. The highest-performing cluster, C1, which includes China, was identified as the group with the most promising investment potential.