DETERMINATION OF IMPORTANT VARIABLES IN DIVORCE TYPE CLASSIFICATION USING THE RANDOM FOREST METHOD WITH SMOTE

Abstract

Central Jakarta is highly strategic area situated at the heart of the Indonesian capital. It serves as the central hub for the government, history, tourism, and elite shopping sectors with convenient access to various buffer areas surrounding the capital. However, the availability of these facilities does not necessarily ensure the continuity of domestic life within the community. This can be observed from the increasing divorce rate in the region since 2017. Notably, a higher proportion of divorce suits are filed by wives than by husbands. There are various factors that can trigger divorce lawsuits such as continuous disputes and arguments, economic factors, and domestic violence. And these factors certainly cannot be separated from the individual profiles of married couples such as age, occupation, education level, and duration of marriage. The purpose of this study is to determine the level of importance of the variables used in the classification of wife-initiated divorce and husband-initiated divorce of married couples in the Central Jakarta area through the Random Forest method. Random Forest is a development of the CART (Classification and Regression Tree) method obtained through the application of bootstrap aggregating and random feature selection methods to the standard CART method. The number of wife-initiated divorce cases exceeds that of husband-initiated divorce cases, necessitating the use of SMOTE technique to address the imbalance in the data set. The results showed that the most important variable used to classify divorce cases was the plaintiff's age followed by the defendant's occupation, the defendant's age, and the plaintiff's occupation.