MQP 2007-Risk Score Assignment for Long-Term Care Underwriting

The project goal was to test a previously designed adaptive risk scoring model for long term care insurance underwriting using actual applicant data from John Hancock. A data filtering method for removing applicants who should not be used to train the model was developed. The model provides accurate risk class assignment, based solely on the medical conditions, when trained on the filtered data. In addition, the model identifies errors in the risk points assigned to individual medical conditions.