The Kangaroo Insurance Case competition organized by Travelers Insurance asked to predict claim costs for auto insurance policies. The 50+ teams were given an anonymized dataset with about 20k training samples. The challenge was to get predictions with the highest gini coefficient on the hold set. Our team consisting of Rajarshi Roychowdhury, Aruni Roychowdhury and Neil Patel ranked second (2nd) in a head-to-head race at the end with a final gini coefficient of 0.229
Business problem You work for Kangaroo Auto Insurance Company, an Australian company. Your business partner, who is not familiar with statistics at all, would like you to create a rating plan based on the historical auto claim data. Your business partner is concerned about segmentation as well as competitiveness, as there are several other competitors in the market. For this case competition, your group’s task is to provide a method for predicting the claim cost for each policy and to convince your business partner that your predictions will work well.
The repo contains all the python notebooks, and sample data.