Enhanced Decision Making in Judicial System of European Court of Human Rights via Bayesian Optimization Boosting Learning

Abstract

As the custodians of constitutional rights and principles, courts bear the responsibility of delivering impartial and equitable judgments. However, the surge in caseloads has necessitated innovative measures to tackle backlog issues, leading to the adoption of automated decision-making processes. This research delves into the implementation of automated decision-making in the European Court of Human Rights (ECHR) Mapping Projects, leveraging publicly available data. The study considers two classes, violation and non-violation, and proposes Bayesian Optimization Boosting Learning (BOBL). A comparative analysis with established machine learning models such as Random Forest (RFT), Gradient Boosting (GBT), Decision Tree (DTE), and Logistic Regression (LRN) is conducted to forecast judicial decisions. The court is segmented into 12 sections, including Court First Section, Court Second Section, Court Third Section, Court Fourth Section, Court Fifth Section, Court Third Section Committee, Court Fifth Section Committee, Court Second Section Committee, Court Grand Chamber, Court Fourth Section Committee, and Court First Section Committee. The results consistently demonstrate BOBL's outperformance over traditional classifiers, showcasing superior accuracy across all court sections and committee scenarios. Noteworthy is BOBL exceptional accuracy, peaking at 99.00%, emphasizing its adaptability and efficacy in diverse legal contexts. These findings suggest that the BOBL model holds significant promise as a resilient and versatile tool for legal applications, surpassing traditional classifiers in both performance and reliability.