Publicação
Understanding and predicting lapses in mortgage life insurance using a machine learning approach
| datacite.subject.fos | Ciências Naturais::Ciências da Computação e da Informação | pt_PT |
| dc.contributor.advisor | António, Nuno Miguel da Conceição | |
| dc.contributor.author | Manteigas, Carlos Manuel Andrade | |
| dc.date.accessioned | 2024-03-14T15:27:04Z | |
| dc.date.available | 2024-03-14T15:27:04Z | |
| dc.date.issued | 2024-02-02 | |
| dc.description | Project Work presented as the partial requirement for obtaining a Master's degree in Data Driven Marketing, specialization in Data Science for Marketing | pt_PT |
| dc.description.abstract | Mortgage life insurance (MLI) offers lucrative opportunities for insurers in Portugal but retaining customers has become challenging amid regulatory changes and fierce competition. After 2009, the market has been reshaped by new competitors with aggressive low premium strategies, posing difficulties for conventional insurers and banks to retain MLI customers. Increasing policy cancellations have become a pressing concern for these established financial institutions. To address this complex and adverse context, this study has tried to develop a predictive model to identify MLI policies prone to lapse and unravel the underlying factors driving this behavior. Using a powerful combination of data from an insurance company and its affiliated bank, the capabilities of four different machine learning (ML) models were explored: Logistic Regression, Random Forest, Neural Networks and XGBoost, with the latter showing the most consistent results. Although the model performs reasonably well, the difficulty in balancing model complexity and generalizability and optimizing Precision and Recall reveals room for improvement. Two novel approaches were introduced by narrowing the focus of the study to one specific insurance protection product, the MLI, and by integrating bank data, to capture multidimensional drivers of lapse behavior, and emphasizing the value of a holistic perspective. Applying SHAP enhanced the interpretability of XGBoost by identifying and explaining the most influential features affecting the predictive model. Notably, the value of external data is underscored as the top four features originated from the bank data. From the insurance company's point of view, this study introduces advanced ML techniques to improve the accuracy of policy lapse prediction, allowing the company to identify and target customers at risk of lapse proactively. | pt_PT |
| dc.identifier.tid | 203544986 | pt_PT |
| dc.identifier.uri | http://hdl.handle.net/10362/164918 | |
| dc.language.iso | eng | pt_PT |
| dc.rights.uri | http://creativecommons.org/licenses/by/4.0/ | pt_PT |
| dc.subject | Mortgage life insurance | pt_PT |
| dc.subject | Lapse risk | pt_PT |
| dc.subject | Machine learning | pt_PT |
| dc.subject | External Data Sources | pt_PT |
| dc.subject | SDG 3 - Good health and well-being | pt_PT |
| dc.subject | SDG 4 - Quality education | pt_PT |
| dc.subject | SDG 9 - Industry, innovation and infrastructure | pt_PT |
| dc.subject | SDG 12 - Responsible production and consumption | pt_PT |
| dc.subject | SDG 13 - Climate action | pt_PT |
| dc.title | Understanding and predicting lapses in mortgage life insurance using a machine learning approach | pt_PT |
| dc.type | master thesis | |
| dspace.entity.type | Publication | |
| rcaap.rights | openAccess | pt_PT |
| rcaap.type | masterThesis | pt_PT |
| thesis.degree.name | Mestrado em Marketing Analítico, especialização em Ciência de Dados Aplicada ao Marketing | pt_PT |