Yüksek Mahkeme Kararlarının Sınıflandırılması için Veri Dengeleme ve Açıklanabilir Yapay Zekâ Tabanlı Karar Destek Sistemi

Artificial intelligence (AI) techniques have great potential in the field of law. Since written texts are mainly used in law, the use of AI techniques can assist decision-making processes by ensuring fast and effective processing of legal texts. In addition, AI techniques can also be used to improve the legal processes and predict the possible outcomes of the cases. In this study, the decisions made by the Court of Cassation and the Council of State, two Supreme Courts in the judicial and administrative judiciary, are considered, and it is aimed to develop an AI-based decision support system. The decisions of the Court of Cassation and the Council of State are crucial for ensuring legal unity, protecting the rights and freedoms of individuals and providing justice. In this context, the decisions of both Supreme Courts taken from the Legislation and Jurisprudence program of the National Judicial Network Project were used, and 11% of these decisions were approvals and 89% were reversals. Therefore, the problem considered is defined as an imbalanced classification problem. First, features were extracted from decision texts using data preprocessing and natural language processing (NLP) techniques. Then, Synthetic Minority Oversampling Technique (SMOTE) and random undersampling were applied to eliminate the imbalance in the data. Finally, in order to predict the court decisions, classification models with k-nearest neighbors, decision tree, support vector machine, random forest, LightGBM, XGBoost and artificial neural networks were developed, and the performances of the models were compared. The experimental results showed that the proposed decision support system has the potential to benefit legal professionals.