Anti-fraud system for detecting non-compliances in corporate spending

Corporate expense reimbursements are common in companies, and these expenses correspond to important amounts as long as the company keeps growing. Thus, the validation and analysis of these transactions are of great importance, intending to verify if they are part of the work’s development and, therefore, should be refunded. The correct identification of this kind of transaction is a complex and expensive task since there is a need for a person to audit each transaction, analyze a macro context and take into account the employee’s role in the expense incurred. The present contribution overcomes these challenges using data analysis and machine learning tools to create a model capable of analyzing and validating a transaction as fraudulent or not. The results for the developed models show a marked improvement over conventional efforts to identify transactions, accurately approaching practical potential levels.