A Self-Attention Network for Hierarchical Data Structures with an Application to Claims Management

TL;DR

This paper introduces a self-attention neural network model tailored for hierarchical healthcare claim data, demonstrating superior performance over traditional and other deep learning models in fraud detection tasks.

Contribution

It presents a novel self-attention based neural network architecture specifically designed for hierarchical, variable-length claim data, outperforming existing models.

Findings

Self-attention model outperforms bag-of-words and CNN models.

Proposed methods achieve higher accuracy on a large claims dataset.

Self-attention model performs best among tested approaches.

Abstract

Insurance companies must manage millions of claims per year. While most of these claims are non-fraudulent, fraud detection is core for insurance companies. The ultimate goal is a predictive model to single out the fraudulent claims and pay out the non-fraudulent ones immediately. Modern machine learning methods are well suited for this kind of problem. Health care claims often have a data structure that is hierarchical and of variable length. We propose one model based on piecewise feed forward neural networks (deep learning) and another model based on self-attention neural networks for the task of claim management. We show that the proposed methods outperform bag-of-words based models, hand designed features, and models based on convolutional neural networks, on a data set of two million health care claims. The proposed self-attention method performs the best.