Aplicação de aprendizado de máquina para classificação de jurisprudências

Abstract

Research and implementation of a web service capable of classifying jurisprudential summaries as to its result, provided or unprovided. Using the python programming language, a supervised machine learning model was developed, which was trained through the use of a predefined jurisprudence base with their respective results. For this, some machine learning algorithms were selected in order to define which one has the best performance: Naive Bayes, Random Forest and K-Nearest Neighbors. After an evaluation of the performance of these algorithms, it was chosen the model based on the Random Forest algorithm, because it has a better performance regarding assertiveness.