Regional Characteristic Analysis of Hometown Love Donation Using k-Means Clustering Algorithm

This study aims to classify South Korean local governments based on data from the first year of the Hometown Love Donation Program in 2023 using machine learning techniques. The primary analytical methods employed are k-means clustering and Principal Component Analysis (PCA). To determine the optimal number of clusters, both the Elbow Method and RMSE of cluster centroids were examined, leading to the selection of 8 clusters as the optimal number. The k-means clustering analysis revealed that the first cluster contains the largest number of local governments, while clusters 2, 5, 7, and 8 include fewer than 10 local governments each. To further analyze cluster characteristics, PCA and PCA-Biplot were performed, with variables normalized using Z-scores for comparison. The PCA results indicated that the first principal component explains most of the data variance, with clusters showing a cross-sectional distribution. PCA-Biplot and Z-score analyses demonstrated that clusters are primarily influenced by variables related to gift items and donation amounts, with the fourth cluster exhibiting predominantly positive values for fund allocation variables. This study provides data-driven insights and institutional recommendations for the Hometown Love Donation Program, distinguishing itself from prior studies through empirical analysis. Limitations include the focus on data from only the first year of the program and a limited range of variables. Future research should consider additional factors such as gift quality, delivery status, donor thank-you notes, and fund utilization to offer a more comprehensive classification of local governments.