Simulation of a Judicial Process using Machine Learning to Analyze Administrative Prejudice and Indicate the Quality of Justice

Abstract

In many countries, it is customary to divide public offenses into different types. For example, in common law countries the doctrine of dividing such offenses into and ""malum prohibitum"" is common. In this article, based on a computer analysis of a large volume of empirical date, we test the following hypotheses, which have essential legal significance. The first hypothesis: judicial decisions rendered under the criminal procedure should have a more developed semantics than judicial decisions rendered under the administrative offence procedure. The second hypothesis: in the case of administrative prejudice, the decision trees built on the materials of criminal proceedings should have a fundamental similarity with the decision trees built on the materials of proceedings on cases of administrative offenses. The study was carried out based on non-political administrative prejudice, which has an important social significance, namely on the example of the failure to pay for the support of children or disabled parents (Article 5.35 of the Administrative Offenses Code of the Russian Federation, Article 157 of the Criminal Code). The mentioned articles of the law are in force in the unchanged version since 15.07.2016, but even before that there was a stable judicial practice on similar cases. The study proved that both hypotheses: judicial decisions rendered under the criminal procedure have a more developed semantics than judicial decisions rendered under the administrative offence procedure. The second hypothesis: in the case of administrative prejudice, the decision trees built on the materials of criminal proceedings demonstrated a fundamental similarity with the decision trees built on the materials of proceedings on cases of administrative offenses.

Keywords

• text mining
• machine learning
• law
• simulation