
Fraud Detection of Travel and Personal Accident Insurance Claims
Implementation Time:
9 months
Solution Provider: AI Singapore
How can Sompo develop a fraud detection application to automatically flag out potential suspicious cases for future investigation, and identify valid claims as candidate for straight-through claim payment?
- – A machine learning solution was developed to process, identify and rank suspicious travel and personal accident claims by giving each claim a fraud score – done daily; results were then submitted for processing by human specialists
- – Models are trained on a smaller, manual labelled data set that is complemented by other data sources e.g. data from internal Sompo claims systems
- – A pipeline was also implemented to allow retraining and updating of models when more data is available
PDF Document
Video link
Solution Diagram
Implementation Time
9 months
Use Case Brochure