In this study, I present the results of my data science internship at Credem bank, focused on the development of machine learning models for the optimization of mortgage application processes. The goal of the internship was to identify patterns and predictors of the duration and the complexity of mortgage applications, in order to streamline and improve the internal management of these processes and eventually provide a real time forecast to the customer. To this end, two machine learning models have been developed: a regressor trained to predict the time remaining for an ongoing mortgage application to reach the deliberation stage, and a classifier trained to classify new mortgage applications based on their complexity. It has been used and tested a combination of supervised and unsupervised techniques to train these models. The results of the analysis showed that both models were able to predict and classify mortgage applications, respectively. For both tasks the ensemble method based on boosting proved to be the most performing ones. Overall, my work at Credem demonstrated the potential of machine learning to improve the efficiency and effectiveness of mortgage application processes in the banking industry.
Improving Mortgage Application Process with Machine Learning: A Case Study at Credem Bank
RINALDI, LUCA
2022/2023
Abstract
In this study, I present the results of my data science internship at Credem bank, focused on the development of machine learning models for the optimization of mortgage application processes. The goal of the internship was to identify patterns and predictors of the duration and the complexity of mortgage applications, in order to streamline and improve the internal management of these processes and eventually provide a real time forecast to the customer. To this end, two machine learning models have been developed: a regressor trained to predict the time remaining for an ongoing mortgage application to reach the deliberation stage, and a classifier trained to classify new mortgage applications based on their complexity. It has been used and tested a combination of supervised and unsupervised techniques to train these models. The results of the analysis showed that both models were able to predict and classify mortgage applications, respectively. For both tasks the ensemble method based on boosting proved to be the most performing ones. Overall, my work at Credem demonstrated the potential of machine learning to improve the efficiency and effectiveness of mortgage application processes in the banking industry.File | Dimensione | Formato | |
---|---|---|---|
Rinaldi_Luca.pdf
accesso riservato
Dimensione
951.24 kB
Formato
Adobe PDF
|
951.24 kB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/45813