Modelo de tomada de decisão aplicado a câmaras colegiadas

Abstract

This work is about the development of a computational model for the decision making of collegiate chambers using the methods of binary logistical regression and decision tree, as to the trial of a fact in collegiate. The model is applicable at the individual vote level, creating data about how different influences (facts characteristics, trial colleagues and from the judge himself) may affect voting in collegiate chambers. Jurisprudence data from a Brazilian tribunal, between 29/09/1997 and 17/11/2011, was used as a case study and the test resulted in the variables Collegiate, Presidentand Type of Defendant showing a bigger influence on the data used for the experiment. The model was applied in sensitivity analysis and simulations using the Net logo environment to test severals cenarios. The percentage of correct guesses that was obtained using the decision tree model in the sample was of 81,33% and for the regression model of 65,4%.