Online Program

Return to main conference page

Thursday, January 11
Thu, Jan 11, 2:00 PM - 3:45 PM
Crystal Ballroom CD
Statistical Integrity of Health Services: Detecting Fraud and Improving Data Quality

The Evolution of Fraud Detection Models (303832)

*Brian Hochrein, IBM Watson Health 

Keywords: Fraud detection; predictive models; prescriptive models; descriptive statistics; social media;

Exploring how fraud detection models have changed over time, we start with descriptive statistics by looking at simple outliers in determining fraudulent activity. We then move to predictive models, supplementing them with social media and network analysis. As the methodology continues to advance we visit Bayesian models along with machine learning techniques. It is important to know whether the sophistication of these models are better at detection or simply keeping up with finding inappropriate behavior.