A predictive performance comparison of machine learning models for judicial cases

Artificial intelligence is currently in the center of attention of legal professionals. In recent years, a variety of efforts have been made to predict judicial decisions using different machine learning models, but no realistic performance comparison between them is available. In this paper, we conducted experiments comparing five well-known machine learning models: Ã-NN, logistic regression, bagging, random forests and SVM. Our experimental results show that the SVM model outperforms the other models over all the different settings, and the semantic information of the text in cases plays an important role in selecting features for the predicting models.