An Innovative AI-Based System for Corruption Risks Assessment Among Corporate Managers to Support Open Source Analysis

Abstract

The paper has its focus on the creation of an innovative Natural Language Processing system for the quest of available information and consequent data analysis, aimed at reconstructing the corporate chain and monitoring the sensitive risk of corruption for people involved in command positions. Today, the greatest opportunity in finding information is represented by the Internet or other open sources, where the contents related to corporate managers are continuously posted and updated. Given the vastness of the information dimension, it seems remarkably advantageous to have an intelligent analysis system capable of independently finding, analyzing and synthesizing information related to a set of target subjects. The aim of this document is to describe a forecasting model based on Machine Learning and Artificial Intelligence techniques capable of understanding whether a news item related to an individual (sought during a due diligence process) contains information about crime, investigation, conviction, fraud, corruption or sanction relating to the subject sought. Methods based on Artificial Neural Networks and Support Vector Machine, compared one to the others, are introduced and applied for the scope. In particular, results showed the architecture based on SVM with TF-IDF matrix and test pre-processing outperforms the others discussed in this paper demonstrating high accuracy and precision in prediction new data as well.