<![CDATA[The increasing number of prison inmates poses a serious challenge for correctional institutions in designing targeted and effective rehabilitation programs. This study aims to compare the performance of K-Means and K-Medoids algorithms in clustering inmate data based on demographic characteristics and crime types. The dataset consists of 2,889 inmate records with seven attributes, including age, gender, religion, sentence duration, crime type, detention status, and other relevant correctional information. The preprocessing stages include feature selection, missing value handling, outlier removal, label encoding, and data standardization using StandardScaler to ensure that all variables are suitable for clustering analysis. The optimal number of clusters was determined using the Elbow Method combined with the Silhouette Score, resulting in K = 3 for both algorithms. The evaluation results show that K-Means outperforms K-Medoids across three internal evaluation metrics, indicating better clustering compactness and separation. However, K-Medoids produces a more balanced and interpretable cluster distribution. Therefore, both methods can support correctional institutions in mapping inmate profiles and formulating more appropriate rehabilitation, supervision, and policy strategies for evidence-based decision making in modern correctional management and inmate development.]]>
Copyrights © 2026
