AI-Powered Risk Control Framework for Health Insurance Fund Management: a Comprehensive Data-Driven Approach

EasyChair Preprint 13719, version 1

Abstract

In the field of AI+ health insurance fund risk control, a comprehensive model leveraging patient data during medical treatment and hospitalization is proposed. The model involves the storage of vast amounts of data, including patient basic information, medication details, surgical procedures, etc., in a data warehouse. Features are then generated based on business indicators, establishing mappings between disease-specific medication and surgical procedures, as well as patient information and medication usage. Risk exposure features are derived from threshold indicators set by the health insurance department's knowledge base, along with outlier analysis for indicators beyond the knowledge base. Unsupervised learning is employed to cluster all medical behaviors, with a focus on clusters with fewer instances to identify and label risks of illegal activities. Subsequently, supervised learning is conducted to classify and quantify illegal risks, providing real-time alerts to healthcare providers and reporting to regulatory authorities for review. Model accuracy is evaluated against audit results, leading to iterative adjustments for enhanced precision and quality of training data.

Keyphrases: AI, Health Insurance, feature engineering, machine learning, model optimization, risk control

@booklet{EasyChair:13719,
  author    = {Pengfei Lin and Yixin Cai and Huasen Wu and Jinghe Yin and Zhaxi Luorang},
  title     = {AI-Powered Risk Control Framework for Health Insurance Fund Management: a Comprehensive Data-Driven Approach},
  howpublished = {EasyChair Preprint 13719},
  year      = {EasyChair, 2024}}