• regression
• prediction
• model
• data mining
• data
• analysis

Decision making is crucial in the bank sector, because banks want to make reliable estimations of profitability in order to minimize non-profitable agreements and the consequent business losses. Indeed, each agreement between the bank, the dealer and the final customer is a configuration based of specific financial conditions.

This thesis, which is the result of an internship project for an Italian bank, investigates the components of the bank's configurations, their outcomes, the bank's predictions about how much profitable an agreement could be. To improve the bank's estimation of margin and profitability, we build, validate, and compare several machine learning predictors (regressors), based on the collected historical data of previous agreements' configuration. Then, we select the best regressor to extract meaningful interpretations and explanations, which help the bank understand the reasons why a given configuration is profitable or not given its financial conditions. The results in this thesis are a first step towards the exploitation of machine learning in the bank's margin and profitability estimation framework.