USO DE MINERAÇÃO DE DADOS PARA AVALIAÇÃO DO DESEMPENHO DA JUSTIÇA ESTADUAL

Abstract

Judicial performance evaluation measures how productive courts are. Data mining techniques allow finding patterns and anomalies given a data set. In this paper we analyzed the productivity of Brazilian state courts through data mining techniques and evaluated the performance of data mining algorithms. The national council of justice annually makes available data on judicial management. Revised studies indicate the main variables used to evaluate the performance of the judiciary, the number of sentences handed down, among others, is the main variable identified as a criterion to evaluate the productivity. Also discussed was the process of extracting knowledge from the database, the Knowledge Discovery in Databases, or KDD. To assess the productivity of the judiciary, the number of judgments handed down by the magistrate was treated, the number of new cases divided by the number of judgments handed down by the magistrate to reflect on the congestion of the courts. For comparison purposes, data were analyzed between 2009 and 2017, grouping them by state and Brazilian region. The performance of the algorithms was measured by the t test. Results obtained from the survey pointed to the most productive state courts, such as the state of Rio de Janeiro, which has a large number of sentences, and the state of Pará, which has one of the fewest congestion numbers. The northeast region has the highest number of judgments by magistrates and the midwest has the least congestion. Despite the difficulty and lack of scientific methodology to measure the performance of the judiciary, the methodologies approached allow good comparisons and a comprehensive discussion.