Insurance Fraud
Insurance Fraud, in all its forms has been estimated to be 33-38% of the total cashflow through the US insurance system. In the UK it is estimated to cost £1.5bn and India and Canada $6.25bn and C$500m respectively. It’s a lot of money however you measure it.
Insurance Fraud Detection
Organised fraud is about industrialised repeat offending. However, inevitably the repetitive nature of the offending will cause patterns in data because keeping every occurrence unique increases the effort and time required for fraudsters. These patterns might be geographic or related to the approach used. These patterns are difficult to see without larger amounts of data and tools to investigate them.
How can Symilarity help identify Insurance Fraud?
Symilarity provides tools to assist with the detection of fraudulent behaviour.
- Helix Enterprise Search/Insight Engine provides functionality to identify fraud within large datasets.
- Intelligent Classification can be trained to identify fraud as each document is being processed.
Repositories
Helix provides a repository based framework to hold data from multiple data sources.
Documents in most common file formats can be ingested. The repository hold an index to the ingested data, making it searchable.
Search
Helix provides full search capability including Boolean operators, fuzzy matching, proximity measures and boosted terms. Helix can search across repositories concurrently (federated search).
Using natural language processing techniques during ingestion, place names can be extracted and enriched with latitude and longitude, so patterns of claims can be searched spatially.
Matching
Helix uses an unsupervised machine learning technique called Vector Space Modelling to create a statistical association between the terms and documents in a corpus. This technique enables documents to be used as search terms, to identify documents that are similar.
Helix can use documents from known fraudulent claims to identify similar documents.
Real Time Classification
Symilarity’s Intelligent Classifier can be trained to identify fraudulent claims. The classifier can be used to classify documents in real time. It operates as an Applications Programming Interface (API), which enables it to be integrated into other processes.
