Comparison and Evaluation of Euclidean Distance and Dice Distance in the K-Means Adaptive Algorithm for Clustering Composite Indexes of Food Security and Vulnerability Maps

This study aims to compare and evaluate the effectiveness of two distance measurement methods, namely Euclidean Distance and Dice Distance, in the K-Means Adaptive algorithm for clustering Food Security and Vulnerability Composite Index data. The dataset used includes index data from 2022 to 2024, comprising 305 entries, which were then cleaned to 298 entries. The evaluation was conducted manually using a sample dataset and automatically using the entire dataset via Google Colab with Python. The algorithm's performance was assessed using the Silhouette Score metric to measure the quality of the resulting clusters. The evaluation results showed that the Euclidean method produced an average Silhouette Score of 0.3082, indicating an suboptimal cluster structure. This study concludes that the choice of distance method significantly influences clustering results, and selection should be tailored to the characteristics of the data.