Solution

Predicting the warranty claims is a regression problem and there are various factors that affect the percentage of claims and these can be classified broadly under customer behavior (end of warranty period) and product behavior (defect information). Few different regression models were tried in R, namely Generalized Linear Models, Generalized Additive models and Random Forest. After evaluating and comparing the results from these 3 models, it was decided that Generalized Additive Models and Random Forest are the two most accurate in predicting the warranty claims percentage and hence an average of both models was used as the final output. In this way, the client was able to detect any issues in the new product at an early stage and was able to fix the problem in the new batches.