Feedback

Faculté des Sciences appliquées
Faculté des Sciences appliquées
MASTER THESIS
VIEW 6 | DOWNLOAD 36

Master Thesis : Continuous learning churn prediction in the context of insurance subscriptions

Download
Poizat, Adrien ULiège
Promotor(s) : Geurts, Pierre ULiège ; Schleich, Corinne
Date of defense : 4-Sep-2023/5-Sep-2023 • Permalink : http://hdl.handle.net/2268.2/18336
Details
Title : Master Thesis : Continuous learning churn prediction in the context of insurance subscriptions
Translated title : [fr] Apprentissage continu et prédiction du churn dans le cadre d'assurances
Author : Poizat, Adrien ULiège
Date of defense  : 4-Sep-2023/5-Sep-2023
Advisor(s) : Geurts, Pierre ULiège
Schleich, Corinne 
Committee's member(s) : Haesbroeck, Gentiane ULiège
Sacré, Pierre ULiège
Language : English
Number of pages : 87
Keywords : [en] machine learning
[en] churn
[en] data science
[en] data analysis
[en] churn prediction
[en] model
[en] Random Forest
[en] Gradient Boosting
[en] XGBoost
[en] On-Line algorithms
[en] imputation
[en] imbalanced data
[en] preprocessing
Discipline(s) : Engineering, computing & technology > Computer science
Funders : NRB Group Belgium
Research unit : Artificial Intelligence Team
Name of the research project : D-predict pipeline
Target public : Researchers
Professionals of domain
Institution(s) : Université de Liège, Liège, Belgique
Degree: Master : ingénieur civil en science des données, à finalité spécialisée
Faculty: Master thesis of the Faculté des Sciences appliquées

Abstract

[en] Customer retention, also known as customer loyalty, refers to the ability of a business to keep its
customers during a specific amount of time, without them leaving for the profit of market competitors. It is a key factor in any business, as several studies have proven that the costs induced by
the set of actions taken to acquire new customers are much higher than the costs to retain existing ones. Besides brand loyalty, several factors can explain why companies are that focused on customer retention. One of the main expenses in the acquisition of new customers revolves around advertising and
the need to have successful promotion campaigns on multiple dimensions. These campaigns sometimes
start without much knowledge, targeting a large audience without much focus. In terms of resources,
finding new customers requires a marketing team that will identify factors and leads. Then, a sales
team should be able to reach out to the targets and convince potential future buyers.
Churn is a phenomenon where customers who have agreed to terms and conditions of a contract
in the past make the choice of canceling or terminating that contract, resulting in the loss of a customer for the company that initiated the contract. In the context of insurance subscriptions, most of
the subscribers who churn are customers who decide not to renew their contract after a given period.
This decision is very challenging to predict, as the cause can come from any source: financial issues,
competitors’ offers, change in lifestyle... These factors are not numerically conceivable as they are specific to each situation. Even if they were, most of them are considered as personal private information
and cannot be gathered legally by companies. Because of that, churn prediction is one of the most
challenging tasks in the field of artificial intelligence.
Throughout this work, the three insurance types that are tackled are Car Insurance, Fire Insurance
(for home tenants) and Fire Insurance (for home owners). Car insurances are mandatory for anyone
and fire insurance for owners. Depending on the conditions of a location, tenants can also be obliged
to subscribe to an insurance too. Besides these insurances, Ethias provides insurance contracts such
as Healthcare, Finance Products, General Assistance or Common Right. The three types for which
we will conduct the study are the three contract types with the highest customer base and, thus, are
of the highest interest for Ethias. It is known that the proportion of customers who churn is very low
in comparison to faithful customers. Such an imbalance can cause severe issues in the classification
process. Binary classification problems exist in multiple fields of work: They are used to diagnose diseases,
for quality control in industries or even spam email detection.The applications of binary classification
are numerous and diverse. In the financial sector, they are applied to trends and market evolution.
Related to finance, in recent days, numerous businesses tend to focus their effort towards churn prediction, as it costs less to keep faithful customers than find leads to acquire new ones. Such a growth in
interest made the companies very dependent on new predicting technologies. In this context, turning
to artificial intelligence has to be considered, as it is a key concept. Machine learning techniques,
especially, whose role is to predict outputs based on a statistical model that combines multiple inputs,
play a major role in these applications.


File(s)

Document(s)

File
Access Master_thesis (9).pdf
Description:
Size: 7.08 MB
Format: Adobe PDF

Annexe(s)

File
Access Specification_SV_pr PhD.xlsx
Description:
Size: 49.54 kB
Format: Microsoft Excel XML

Author

  • Poizat, Adrien ULiège Université de Liège > Master ingé. civ. sc. don. à . fin.

Promotor(s)

Committee's member(s)

  • Haesbroeck, Gentiane ULiège Université de Liège - ULiège > Département de mathématique > Doyenne de la Faculté des Sciences
    ORBi View his publications on ORBi
  • Sacré, Pierre ULiège Université de Liège - ULiège > Dép. d'électric., électron. et informat. (Inst.Montefiore) > Robotique intelligente
    ORBi View his publications on ORBi
  • Total number of views 6
  • Total number of downloads 36










All documents available on MatheO are protected by copyright and subject to the usual rules for fair use.
The University of Liège does not guarantee the scientific quality of these students' works or the accuracy of all the information they contain.