IMPLEMENTATION OF LATENT DIRICHLET ALLOCATION (LDA) FOR DIVORCE FACTOR TOPIC MODELING (Case Study: Divorce Decision Document at the Yogyakarta Religious Court in 2020) Elen Prihana Putri, Ayundyah Kesumawati
Statistics Department of Universitas Islam Indonesia
Abstract
Marriage is one of the acts of worship that must be properly maintained in order to achieve an eternal bond between a man and a woman as husband and wife with the aim of forming a sakinah, mawaddah, warahmah family. However, often the goals of the marriage cannot be realized properly and run aground in the middle of life, and quarrels often occur until a divorce is created. The phenomenon of divorce cannot be separated from various causal factors, so that it becomes a reason for husbands or wives to file for divorce. The factors that cause divorce are contained in the divorce case judge^s decision document. The aim of this research is to implement topic modelling using the Latent Dirichlet Allocation (LDA) method in the judge^s decision documents so that several topics related to the causes of divorce are obtained. This research focuses on the decision documents of divorce case judges at the Yogyakarta Religious Court in 2020. From the results of the descriptive analysis, most of the divorce cases were contested with a percentage of 78% compared to talak divorce, then the most dominant cause of divorce is disputes and constant fights followed by leaving other parties, the economy, apostasy, madat, domestic violence, and disability. Then the results of the topic modelling analysis using the LDA method obtained a total of 3 topics with a coherence value of 0,45608. Model topic 1 tends to discuss economic problems, model topic 2 tends to discuss domestic disputes and model topic 3 tends to discuss the existence of a third party.
