Incorporating Machine Learning in Dispute Resolution and Settlement Process for Financial Fraud

Abstract

Abstract This paper aims to classify disciplinary hearings into two types (settlement and contested). The objective is to employ binary machine learning classifier algorithms to predict the hearing outcomes given a set of features representing the victims, offenders, and enforcement. Data for this project came from the Investment Industry Regulatory Industry of Canada’s (IIROC) tribunal hearing. The data comprises cases that made their way through the IIROC ethics enforcement system and were decided or negotiated by a hearing panel. The findings from the machine learning classifiers confirm that decisions in these cases are not proportionate to the harm committed and that the presence of aggravating factors does not result in harsher sentences.