Penerapan Algoritma C4.5 Untuk Klasifikasi Kasus Perceraian Di Indonesia

Abstract

Divorce cases in Indonesia continue to show a significant increase. Based on data from Statistics Indonesia (BPS) in 2024, hundreds of thousands of divorce cases were recorded across all provinces. This study applies the C4.5 algorithm to classify the level of divorce cases using a BPS dataset covering 34 provinces, consisting of 21 regular attributes and 2 specific attributes representing the contributing factors of divorce at the regional level. The model was built using cross-validation and pruning mechanisms to reduce the risk of overfitting. The evaluation results show that the model achieved an accuracy of 85.29%, with weighted mean recall and weighted mean precision in the range of 84–85%, indicating a relatively stable classification performance, although not yet optimal. The confusion matrix shows that most predictions are accurate, but there are still misclassifications, particularly between the Medium and High categories. The final decision tree reveals that the “Cerai Gugat” (lawsuit-initiated divorce) attribute serves as the main splitter at the root node, followed by the attribute “Continuous Conflicts and Disputes” at the subsequent level. This indicates that these two attributes have dominant contributions in determining the classification category of divorce cases. Overall, this study demonstrates that the C4.5 algorithm is quite effective in identifying divorce patterns based on BPS data and has the potential to support the development of decision-support systems for government institutions.