Data Science Forensics
Data science forensics is a hot growing field for professional data scientists. Usually employed to detect financial and business fraud, it is now being used to detect voting fraud in the US. Teaming with data engineers, computer scientists, lawyers, law enforcement and regulators, forensic data scientists look for anomalies to flag for further investigation.
Using anomaly detection algorithms and high performance compute on massive data sets saves time and brainpower. Anomalies can be detected showing statistically improbable outcomes using algorithms employing techniques like Fourier Spectra Histograms, Newcomb–Benford Law, Last Digit Frequency Analysis, Chi Squared Goodness of Fit Test, and other anomaly detection techniques. While anomalies themselves are not necessarily fraud, they can help detect evidence of fraud.
In the case of voting, examining times series logs of ballot counting processes at precinct, state and national levels may show statistical anomalies that indicate narrowly targeted fraud or broader systemic fraud. In addition, an audit can compare paper ballot recounts to voter machine tabulated tallies potentially showing significant discrepancies that may constitute evidence of fraud. Moreover, examination of voting machine software, algorithms and human access logs can determine changes, potential security breaches and tampering. This can lead to discovery of hard evidence of voting fraud.
Data science can play an important forensic role in detecting fraud in a myriad of fields.