Insurance Fraud Detection with Unsupervised Deep Learning

报告摘要：In this talk, I will present a deep learning methodology to gain pragmatic insights into the behavior of an insured person using unsupervised variable importance. Starting with a preliminary investigation of the limitations of the existing fraud detection models, we propose a variable importance methodology incorporated with two prominent unsupervised deep learning models, namely, the autoencoder and the variational autoencoder. Both qualitative and quantitative performance evaluations are conducted. Real data sets are used to illustrate the idea and methodology. I will mention some related recent work at the end of the talk.