Life Insurance Prediction and Its Sustainability Using Machine Learning Approach

Owning life insurance coverage that is not enough to pay for the expenses is called underinsurance, and it has been found to have a significant influence on the sustainability and financial health of families. However, insurance companies need to have a good profile of potential policyholders. Customer profiling has become one of the essential marketing strategies for any sustainable business, such as the insurance market, to identify potential life insurance purchasers. One well-known method of carrying out customer profiling and segmenting is machine learning. Hence, this study aims to provide a helpful framework for predicting potential life insurance policyholders using a data mining approach with different sampling methods and to lead to a transition to sustainable life insurance industry development. Various samplings, such as the Synthetic Minority Over-sampling Technique, Randomly Under-Sampling, and ensemble (bagging and boosting) techniques, are proposed to handle the imbalanced dataset. The result reveals that the decision tree is the best performer according to ROC and, according to balanced accuracy, F-1 score, and GM comparison, Naive Bayes seems to be the best performer. It is also found that ensemble models do not guarantee high performance in this imbalanced dataset. However, the ensembled and sampling method plays a significant role in overcoming the imbalanced problem.