Predictive churn models in vehicle insurance
| Autor(a) principal: | |
|---|---|
| Data de Publicação: | 2019 |
| Tipo de documento: | Dissertação |
| Idioma: | eng |
| Título da fonte: | Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
| Texto Completo: | http://hdl.handle.net/10362/90767 |
Resumo: | Internship Report presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics |
| id |
RCAP_31cb881cda0f942a9385390396a2eadc |
|---|---|
| oai_identifier_str |
oai:run.unl.pt:10362/90767 |
| network_acronym_str |
RCAP |
| network_name_str |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
| repository_id_str |
7160 |
| spelling |
Predictive churn models in vehicle insuranceSupervised learningPredictive churn modelsMachine LearningEnsembleInternship Report presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced AnalyticsThe goal of this project is to develop a predictive model to reduce customer churn from a company. In order to reduce churn, the model will identify customers who may be thinking of ending their patronage. The model also seeks to identify the reasons behind the customers decision to leave, to enable the company to take appropriate counter measures. The company in question is an insurance company in Portugal, Tranquilidade, and this project will focus in particular on their vehicle insurance products. Customer churn will be calculated in relation to two insurance policies; the compulsory motor’s (third party liability) policy and the optional Kasko’s (first party liability) policy. This model will use information the company holds internally on their customers, as well as commercial, vehicle, policy details and external information (from census). The first step of the analysis was data pre-processing with data cleaning, transformation and reduction (especially, for redundancy); in particular, concept hierarchy generation was performed for nominal data. As the percentage of churn is not comparable with the active policy products, the dataset is unbalanced. In order to resolve this an under-sampling technique was used. To force the models to learn how to identify the churn cases, samples of the majority class were separated in such a way as to balance with the minority class. To prevent any loss of information, all the samples of the majority class were studied with the minority class. The predictive models used are generalized linear models, random forests and artificial neural networks, parameter tuning was also conducted. A further validation was also performed on a recent new sample, without any data leakage. In relation to compulsory motor’s insurances, the recommended model is an artificial neural network. The model has a first layer of 15 neurons and a second layer of 4 neurons, with an AUC of 68.72%, a sensitivity of 33.14% and a precision of 27%. For the Kasko’s insurances, the suggested model is a random forest with 325 decision trees with an AUC of 72.58%, a sensitivity of 36.85% and a precision of 31.70%. AUCs are aligned with other predictive churn model results, however, precision and sensitivity measures are worse than in telecommunication churn models’, but comparable with insurance churn predictions. Not only do the models allow for the creation of a churn classification, but they are also able to give some insight about this phenomenon, and therefore provide useful information and data which the company can use and analyze in order to reduce the customer churn rate. However, there are some hidden factors that couldn’t be accounted for with the information available, such as; competitors’ market and client interaction, if these could be integrated a better prediction could be achieved.Vanneschi, LeonardoMendes, Jorge MoraisRUNBellani, Carolina2020-01-06T19:32:25Z2019-11-072019-11-07T00:00:00Zinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesisapplication/pdfhttp://hdl.handle.net/10362/90767TID:202358011enginfo:eu-repo/semantics/openAccessreponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãoinstacron:RCAAP2024-03-11T04:40:17Zoai:run.unl.pt:10362/90767Portal AgregadorONGhttps://www.rcaap.pt/oai/openaireopendoar:71602024-03-20T03:37:10.733501Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãofalse |
| dc.title.none.fl_str_mv |
Predictive churn models in vehicle insurance |
| title |
Predictive churn models in vehicle insurance |
| spellingShingle |
Predictive churn models in vehicle insurance Bellani, Carolina Supervised learning Predictive churn models Machine Learning Ensemble |
| title_short |
Predictive churn models in vehicle insurance |
| title_full |
Predictive churn models in vehicle insurance |
| title_fullStr |
Predictive churn models in vehicle insurance |
| title_full_unstemmed |
Predictive churn models in vehicle insurance |
| title_sort |
Predictive churn models in vehicle insurance |
| author |
Bellani, Carolina |
| author_facet |
Bellani, Carolina |
| author_role |
author |
| dc.contributor.none.fl_str_mv |
Vanneschi, Leonardo Mendes, Jorge Morais RUN |
| dc.contributor.author.fl_str_mv |
Bellani, Carolina |
| dc.subject.por.fl_str_mv |
Supervised learning Predictive churn models Machine Learning Ensemble |
| topic |
Supervised learning Predictive churn models Machine Learning Ensemble |
| description |
Internship Report presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics |
| publishDate |
2019 |
| dc.date.none.fl_str_mv |
2019-11-07 2019-11-07T00:00:00Z 2020-01-06T19:32:25Z |
| dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
| dc.type.driver.fl_str_mv |
info:eu-repo/semantics/masterThesis |
| format |
masterThesis |
| status_str |
publishedVersion |
| dc.identifier.uri.fl_str_mv |
http://hdl.handle.net/10362/90767 TID:202358011 |
| url |
http://hdl.handle.net/10362/90767 |
| identifier_str_mv |
TID:202358011 |
| dc.language.iso.fl_str_mv |
eng |
| language |
eng |
| dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
| eu_rights_str_mv |
openAccess |
| dc.format.none.fl_str_mv |
application/pdf |
| dc.source.none.fl_str_mv |
reponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação instacron:RCAAP |
| instname_str |
Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação |
| instacron_str |
RCAAP |
| institution |
RCAAP |
| reponame_str |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
| collection |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
| repository.name.fl_str_mv |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação |
| repository.mail.fl_str_mv |
|
| _version_ |
1799137988962156544 |