MQP 2010-Predicting Policyholder Behavior and Benefit Utilization: An Analysis on Long-Term Care Insurance

In order to better serve their customers, a project to create a methodology for identifying variables that could indicate future long-term care insurance usage was commissioned by Ability Resources, Inc. As a basis for constructing a predictive model, tools such as SAS and Excel were implemented. A k-means clustering algorithm in SAS was utilized to group policyholders with similar characteristics, and a performance evaluation was executed in Excel. Together, these processes created a tool that determined the impact each characteristic had on policyholder benefit utilization. The validity of the process was assessed by applying it to supplemental data generated by the team. After several trials, the Variable Identification Procedure proved accurate.