Insurance fraud costs the economy over $2 billion per annum – our ML model helped a leading insurance provider to detect fraudulent claims
What we were solving for
Quality of detection processes: Insurance companies struggle to detect fraudulent claims, given the use of sophisticated methods to conceal fraud and quantum of data to analyse
Resourcing: Assessing and processing claims is a time-consuming and labor-intensive process and requires a significant investment in technology and personnel. Furthermore, insurance companies must comply with a wide range of regulations related to claims handling and processing
OptiSol solution
Created an unsupervised learning model that would consume data provided in an insurance claim and then flag if suspicious – a classification model, trained and tested with historical data
Isolation Forest and SVM (Support Vector Machine) model was constructed to identify claims that stood apart as anomalous
Data is analysed to examine and check for abnormal data points, as well as to comprehend variables that may cause anomalies. e.g., the duration between the accident date and the claim reported date
Driver specific analysis is automated in which the license class, experience and vehicle-based restrictions are considered