Abstract— The application of data mining technique
currently is widely used in supporting business activity,
especially for insurance company which has to analyze a big
number of customer data. The insurance company has to
predict its new customer acquisition as well as maintain its
existing customers. This paper is focused on how we support
insurance companies, especially PT. XYZ, to analyze their
customers’ characteristics data using the best data mining
algorithm technique. The research aims to analyze existing
customer data and to predict as well as find optimal patterns of
how many of their customers will extend their insurance
policies, and how many will not. We also explore the customer
retention rate discovering the optimal solution for the company.
We applied 4 different algorithms were applied, i.e. support
vector machine algorithm, decision tree, k-NN, and random
forest algorithm, comparing the results and finding a better
solution. From the analysis, we found that the random forest
algorithm provides better results in predicting the status of the
insurance policy extension of current customers, with an
accuracy rate of 91.08% and AUC value of 0.962. This result is
quite good for PT XYZ, and could be enhanced in the future by
applying a good strategy to increase their customer renewal
ratio.