Abstract:
The thesis aims at predicting the probability of default in the risk management framework.
The work, structured in two main sections, looks at how machine learning techniques might be used to model credit risk, given a dataset of consumers of several European financial institutions.
The first part of the thesis will be focused on the analysis of different features of the dataset, with the aim of trying to identify if there are some patterns that link them to the customers' probability of default.
In the second section, it will be approached the empirical part of the work, in which will be developed different machine learning models, in order to estimate the probability of default.
In the end, will be presented the results obtained by each prediction model, and, according to different metrics, will be determined which of them is the more suitable for the prediction purpose.
