Analysis and Clustering of Plantation Crop Production in Indonesia using K-Means Clustering and Principal Component Analysis Methods

Abstract

Indonesia has substantial potential in the plantation sector, with production levels varying considerably across provinces. These differences are influenced by geographical conditions, land availability, and the leading commodities of each region. This study aims to identify production patterns of plantation crops in Indonesia using K-Means Clustering and to visualize the resulting clusters using Principal Component Analysis (PCA). Official 2024 BPS data covering the production of oil palm, coconut, rubber, coffee, cocoa, tea, and sugarcane across 38 provinces were analyzed. The preprocessing procedure included outlier treatment using Interquartile Range (IQR)-based capping, logarithmic transformation, normalization using StandardScaler, dimensionality reduction using PCA, and clustering using the K-Means algorithm. The optimal number of clusters was determined using the Elbow Method, Silhouette Score, and Davies-Bouldin Index (DBI). The results indicate that K = 5 was the optimal clustering solution, achieving a Silhouette Score of 0.5883 and a DBI of 0.4744. The clustering results classified Indonesian provinces into five production categories: very high, high, moderate, low, and very low. Provinces in Sumatra and Kalimantan dominated the very-high and high production categories, while several provinces fell into the moderate category. In contrast, most provinces in eastern Indonesia were classified into the low and very-low production categories. Furthermore, the combination of K-Means Clustering and PCA proved effective for categorizing provincial production data and identifying spatial patterns in plantation crop production across Indonesia. The findings can support government agencies and other stakeholders in formulating plantation-sector development policies and establishing regional priorities based on the production characteristics of individual provinces.