Predicting court motion outcomes · EL GATO Lab

With Dr. Derek Aguiar. Civil-court motions have rich text (briefs) and tabular structure (counts, parties, dates). I built features for both and combined them.

Highlights

• Beta-Bernoulli / one-hot for tabular; TF-IDF + Word2Vec for documents.
• Decision Tree, XGBoost, SVM for up to 65% prediction accuracy.
• Anomaly-detection methods (Isolation Forest, One-Class SVM) for data filtering.