Recommender Systems for Insurance Marketing

Recommender systems are machine learning algorithms typically employed to support marketing decisions, as they identify statistically validated associations between products and consumers. These tools have been successfully adopted in many fields; however, not much has been done for the insurance industry. The paper aims to fill that gap by giving an overview of various relevant recommender systems, both from a theoretical and a practical perspective. In particular, classical approaches are compared to methods based on machine learning techniques. Finally, we present a comparison of the performance of the reviewed methodologies in the context of an insurance case study. It appears that machine learning–based approaches (such as gradient boosting trees and deep neural networks) generally provide stronger performance at the cost of longer computational times. Address for Correspondence: spedicato_giorgio@yahoo.it