Abstract
Divorce is a social phenomenon whose rate continues to rise in various regions of Indonesia and has become a shared concern for both the government and society, since it may give rise to social, psychological, and economic consequences for the individuals and families involved. Divorce is generally caused by several interrelated factors, namely economic problems, continuous domestic disputes, infidelity, and domestic violence (KDRT). This study aims to apply the K-Nearest Neighbor (KNN) algorithm to classify the level of divorce (Low, Medium, and High) based on these four causal factors. The data used are secondary data obtained from the Central Bureau of Statistics (BPS) covering ten provinces on Sumatra Island for the 2023-2025 period. Prior to classification, the data were normalized using the Min-Max Normalization method so that every attribute shared a comparable value range. The dataset was divided into 20 training data (2023-2024) and 10 testing data (2025), with the value of K determined using the formula K = √n, resulting in K = 5. The classification process was carried out by calculating the Euclidean Distance between the testing data and the entire training data, then determining the predicted class through majority voting among the five nearest neighbors. The classification model was implemented in a web-based application built with PHP and MySQL that supports dataset import, data normalization, KNN computation, and classification result display. The test results on the ten testing data for 2025 show that the system achieved an overall accuracy of 70%, with the best performance obtained for the Low class (recall of 100%), while the Medium and High classes still experienced misclassification due to the limited amount of training data in those classes. These results indicate that the KNN algorithm is capable of recognizing divorce patterns based on their causal factors and can be used as a decision-support tool related to the analysis of divorce rates on Sumatra Island.
