EXECUTIVE SUMMARY

By embedding a combination of graph databases, machine learning and data visualization within Claims Investigation Units we developed a solution that prioritised claims for investigation, improved fraud detection rates.

OUTLINE

Working with the team responsible for claims handling, we implemented fraud solutions using data visualisations, graph databases and custom developed machine learning algorithms.  The solution allowed for the timely intervention of claims by CIU agents. In the first year of operation this increased fraud savings by 25%, amounting to approximately €4 million and allowed easier and faster investigation of claims.

SOLUTION

A graph database was built using entities such as incident/claim, car, claimant details, and other data, along with previous claims deemed suspicious.  Using interactive data visualisation tools CIU members investigated claims faster and checked for suspicious behaviours and patterns. Through the implementation of an alert system based upon either known fraud heuristics and the outputs of a machine learning model, CIU investigators were notified of problematic claims, enabling a prioritised response.

BUSINESS OUTCOMES

  • Increased fraud savings.
  • Claims investigated quicker at lower cost.
  • Custom visualisation tools to investigate complex data.
  • Easier investigation and management of problematic claims.

KEY BENEFITS

  • Reduction in claims and subsequent costs in handling problematic claims
  • Empowerment of CIU investigators allowing greater efficiencies in claims investigation
  • reduction in risk of missing problematic claims

TOOLS AND TECHNIQUES

  • i2, iBASE, iBASE designer, i2 Analyst Notebook,
  • SQLServer,
  • R,
  • Machine Learning.

CONTACT DETAILS

Michael Crawford
M: +353 (0)87 996 7437
E: mcrawford@describedata.com