Application of the K-Means Algorithm for Clustering the Characteristics of Customs Valuation Objections in the Siap Tanding Application

Abstract

Customs valuation objections represent an important aspect of customs administration that requires systematic analysis to understand the characteristics and distribution of cases. The availability of administrative records through the SIAP TANDING application creates opportunities to identify patterns using data mining techniques. This study aims to apply the K-Means clustering algorithm to group customs valuation objection records according to similarities in their relevant attributes. The research methodology consists of data collection, data preprocessing, feature selection, data normalization, K-Means implementation, and cluster evaluation. The dataset is planned to comprise 500 customs valuation objection records, with attributes selected according to data availability, including customs value, adjustment amount, objection value, and case resolution duration where applicable. The Elbow method is used to determine an appropriate number of clusters, while the Silhouette Score is employed to evaluate cluster cohesion and separation. The resulting cluster profiles are intended to reveal differences in the characteristics of customs valuation objection cases and provide a structured understanding of patterns within the administrative data. The proposed approach demonstrates the potential of unsupervised machine learning to support data-driven analysis in customs administration. The findings are expected to provide additional analytical information for evaluating case characteristics and supporting decision-making in the SIAP TANDING context. Further validation using the actual dataset is necessary to establish the quality, reliability, and practical relevance of the resulting clusters.