Machine learning based approaches for intelligent adaptation and prediction in banking business processes. (c2018)

Abstract

Companies, nowadays, rely on systems and applications to automate their business processes and data management. In this context, the notion of integrating machine learning techniques in banking business processes has emerged, where trainable computational algorithms can be improved by learning. Our objective in this work is to propose machine learning models that can benefit from the historical data available in banking environment in order to improve and automate their business processes. In this context, we first propose in this thesis a model providing Intelligent Behavior- Aware Adaptation of Roles using Machine Learning Classification. The proposed scheme is capable of assessing the deployed access control polices and updating them systematically with new roles based on employees behaviors and system constraints. Experiments on real life data set explore the feasibility of our approach, which also provides better performance in terms of required authorizations, transactions time and employees working hours. Moreover, we propose in this thesis a Deep Learning Based Approach to Predict Non-Performing Loans. Compared to the literature, the proposed model embeds a new feature selection method and offers higher detection accuracy, which helps lenders and financial institutions to better manage their lending activities and loan monitoring processes.