Benford’s Law and Artificial Intelligence Applied to COVID-19

Abstract

Newcomb-Benford Law or Benford Law (BL) is a simple and powerful tool to identifying potencial anomalies in supposedly natural phenomena. BL works by comparing the frequency of the first digits acquired from an event with a pattern empirically established by Benford. The behavior described by Benford is typical in many natural processes and, therefore, several studies use the technique to try to identify anomalies that might suggest fraud in some data sets. Another trend is the use of tools that use artificial intelligence to support auditing. Considering that a COVID-19 pandemic is a natural event, it is possible to establish criteria for comparing the numbers released by governments and their relationship with BL. This research models Support Vector Machines (SVM) according to BL and makes a reliability analysis of the numbers of new cases and deaths, considering the pandemic scenario in 11 countries. Then, the work makes a statistical analysis according to BL and compares it to the results predicted by the algorithm. The results show that the network was able to make predictions that reinforce the BL results. Only two countries (Germany and Japan) presented results fully adherent to BL, either by statistical treatment or SVM prediction in all scenarios. The article used the data provided by Johns Hopkins University.