<![CDATA[This study aims to systematically analyze the application of Artificial Intelligence and Big Data in legal case classification and judicial decision prediction. The study employs a Systematic Literature Review by examining relevant publications on the use of artificial intelligence technologies in the legal domain. The review focuses on the methods employed, dataset types and characteristics, preprocessing techniques, feature extraction methods, and factors affecting model performance. The findings show that methods used for case classification and judicial decision prediction include Machine Learning, Natural Language Processing, Support Vector Machine, Deep Learning, BERT, LSTM-CNN, Contrastive Learning, Knowledge Graph, Graph Contrastive Learning, Large Language Model, Multi-Task Learning, AdaBoost, and Explainable Deep Learning. The datasets are predominantly composed of legal documents, case texts, judgment texts, case facts, appeal cases, and structured judicial data. Data representation is performed through NLP, BERT, Deep Learning models, Fact Elements, Knowledge Graphs, and graph-based approaches. Preprocessing techniques are not reported consistently across the reviewed literature. Model performance is influenced by dataset quality and characteristics, data representation, algorithm and model architecture selection, completeness of case facts, relationships among legal elements, and the ability of models to capture legal context. The findings indicate a methodological development from Machine Learning and NLP toward Deep Learning, large language models, and graph-based approaches for supporting case analysis and judicial decision prediction. The review also emphasizes interpretability for responsible legal analysis and judicial decision support systems.]]>
