Abstract
This study explores the use of machine learning to predict judicial decisions in criminal cases from the Oromia Supreme Court. A dataset of 1638 cases was collected and pre-processed, and various ML models were applied with different feature extraction techniques. The Random Forest model with TF-IDF features achieved the highest accuracy for judgment prediction (98.5%), while the Support Vector Machine model with TF-IDF features performed best for penalty prediction (79.68%). Legal experts confirmed the model's effectiveness with a 77.5% accuracy rate. This study highlights the potential of ML for predicting judicial outcomes in criminal cases and recommends further exploration for potential implementation in court systems.