On identifying early blockable taxpayers on goods and services trading operations.

Goods and services trading taxation faces a series of challenges around the world, either by its decentralized and adaptive aspect, or by the recurrent and increasingly elaborate fraudsters' attempts seeking tax evasion. Machine learning techniques emerge as a powerful tool for analyzing a large volume of data in an effective and agile way, allowing to preemptively identify and stop suspicious behavior before fraudsters cause more effective damage to public treasury. In this sense, this work presents an analysis of classifying algorithms for identifying taxpayers whose suspicious transactions indicate a possible issuer or receiver of fictitious invoices, leading to its blockage and consequently interrupting its activities. The results, analyzed in comparison with the currently executed manual process, show how relevant the gains are when this resource is added to it.