Prediction of Customer Underwriting of Policies in Banking Institutions Through Machine Learning

Policies are important for banks because they provide them with liquidity and make it easier for them to know in advance how much money they have available to carry out their different financial activities. For this reason, a challenge for banks is to optimize their marketing campaigns with which they offer policies to their customers, and in order to contribute to this problem, this article has posed the challenge of predicting whether or not a customer will subscribe to the policy; thus optimizing the marketing campaign since this financial product could be offered only to potential buyers. The CRISP-DM methodology has been used by structuring it in 3 phases: data collection and extraction, data preparation, and finally modeling or prediction; which allows predicting with a high percentage of accuracy if the customer would subscribe or not. To demonstrate the effectiveness of our method, we use the public data set Bank Marketing Data Set that has a large number of customers with characteristics such as age, marital status, type of work, level of education, whether they own a home, among others, and we use quality measures for classification. This opens the door for banks to predict their potential customers for policies, as well as their liquidity and grant more loans to companies in need, and how future work can include additional data preparation processes.