USING MACHINE LEARNING TO PREDICT PERFORMANCE OF TRIAL COURT ADMINISTRATION: AN EMPIRICAL STUDY WITH IRANIAN PERFORMANCE INDICATORS OF TRIAL CASE PROCESSING

Abstract

This paper explores the significance of evaluating justice system performance to ensure effectiveness across diverse legal frameworks. Traditionally, methods like expert surveys and document analysis were used to generate empirical indicators. However, this study employs machine learning to predict trial court performance, using key processing indicators. Data from 21 branches of the General Court of Law in Tehran, Iran, comprising 119 case management records, is analyzed. logistic regression proves most effective among various models, achieving 98.5% AUC and 95.0% CA. Results indicate that resolved cases impact positively, while pending cases have minimal influence. Monitoring time and working days contribute insignificantly. Early detection of negative performance issues is crucial for maintaining public trust. Regular evaluations not only enhance court efficiency but also aid in developing decision support systems for improved performance.