Detecting Claims Fraud Using Text Data and Bi-Directional LSTM Networks With Attention
According to the Insurance Information Institute, fraud accounts for about 10 percent of incurred losses and loss adjustment expenses, costing property-casualty insurers an estimated $30 billion each year. Ultimately, this cost is passed on to the consumer in the form of higher insurance premiums. The focus of this presentation will be to discuss how bi-directional long-short term memory (LSTM) networks with attention can be applied to claims textual data, to flag and reduce claims fraud. I will discuss the business context, technical solution, as well as ongoing and future work.
Peng Lee is a Senior Data Scientist at American Family Insurance with over 13 years of industry experience. He is an accomplished machine learning scientist with a track record of building and deploying models that directly impact the company’s bottom line. He is passionate about understanding the latest machine learning research and how it can be leveraged. He holds a Bachelor of Mathematics from University of Wisconsin Madison and a Master of Predictive Analytics from Northwestern University. He is a Fellow of the Casualty Actuarial Society.