A recommendation system for car insurance

  • PDF / 1,895,311 Bytes
  • 22 Pages / 439.37 x 666.142 pts Page_size
  • 32 Downloads / 369 Views

DOWNLOAD

REPORT


A recommendation system for car insurance Laurent Lesage1,2   · Madalina Deaconu2 · Antoine Lejay2 · Jorge Augusto Meira1 · Geoffrey Nichil3 · Radu State1 Received: 25 November 2019 / Revised: 3 March 2020 / Accepted: 14 May 2020 © EAJ Association 2020

Abstract We construct a recommendation system for car insurance, to allow agents to optimize up-selling performances, by selecting customers who are most likely to subscribe an additional cover. The originality of our recommendation system is to be suited for the insurance context. While traditional recommendation systems, designed for online platforms (e.g. e-commerce, videos), are constructed on huge datasets and aim to suggest the next best offer, insurance products have specific properties which imply that we must adopt a different approach. Our recommendation system combines the XGBoost algorithm and the Apriori algorithm to choose which customer should be recommended and which cover to recommend, respectively. It has been tested in a pilot phase of around 150 recommendations, which shows that the approach outperforms standard results for similar up-selling campaigns. Keywords  Recommendation system · Up-selling · Car insurance · XGBoost algorithm · Apriori algorithm

1 Summary Global purpose In this paper, we propose a recommendation system built for a better customers’ experience, by suggesting them the most appropriate cover in time. The requirement for this system is to perform a more efficient up-selling than classic marketing campaigns. Recently, the applicability of machine learning algorithms have become very popular in many different areas of knowledge leading to learn up-to-date advanced patterns from customers’ behaviour and consequently target

* Laurent Lesage [email protected] 1

University of Luxembourg, 29, Avenue JF Kennedy JFK Building, E02‑206, 1855 Luxembourg, Luxembourg

2

Université de Lorraine CNRS, Inria, IECL, 54000 Nancy, France

3

Foyer Assurances, 12, Rue Léon Laval, 3372 Leudelange, Luxembourg



13

Vol.:(0123456789)



L. Lesage et al.

customers more accurately. In the context of recommendation systems, such algorithms generate automatically commercial opportunities suited to each customer. Purpose: up-selling Our recommendation system is currently in use by Foyer Assurances1 agents. Our goal is to support the agents that are and will continue to be the best advisers for customers, due to their experience and their knowledge of their portfolio. In short, our tool helps them by automatically selecting from their large portfolios the customers most likely to augment their insurance coverage, in order to optimize up-selling campaigns for instance. Thus, an insurance company using this solution could combine advantages from both data analysis and human expertise. Agents validate if the recommendations from our system are appropriate to customers and make trustworthy commercial opportunities for them. The recommendation system is also planned to be integrated in customers’ web-pages, in order to provide them a personalized as