Analisis Prediktif Putusan Perdata Berbasis Machine learning di Indonesia

The rapid advancement of artificial intelligence, particularly machine learning (ML), has opened new opportunities in the legal domain, especially in addressing the long-standing issue of inconsistency in civil court decisions in Indonesia. This study aims to develop and evaluate predictive models of civil case outcomes using various ML approaches, including Logistic Regression, Support Vector Machine, Random Forest, XGBoost, and IndoBERT. A dataset of 199,000 published civil court decisions was collected, pre-processed, and annotated into three categories: granted, rejected, and partially granted. The experimental results demonstrate that IndoBERT achieved the best performance with an accuracy of 83.5% and an F1-macro score of 81.7%, outperforming classical models. Feature analysis indicated that contractual terms, evidence, and core legal reasoning were the most influential predictors. These findings highlight the potential of ML to enhance consistency, transparency, and predictability in the Indonesian judiciary, while also raising important considerations regarding ethics, bias, and interpretability. The study contributes to both the theoretical discourse on legal analytics and the practical implementation of AI in judicial reform.