Machine learning models for default propagation among the supply network of Italian corporates.

The European Central Bank requires banks to adapt their capabilities in terms of organizations, Process and IT infrastructure as a suite of integrated answers to the non-performing loans problem. In this regard Intesa Sanpaolo Group (henceforth "ISP") dedicated a specific section in the 2018-2021 business plan highlighting the intention to improve the proactive management of the loan portfolio, with the aim of significantly reducing the level of probable defaults of ISP and optimizing the risk-return profile. At the end of 2016 ISP launched the Early Warning System tool to strengthen and evolve the interception of signals of credit deterioration from several indicators, particularly relevant for a timely detection of the transition to distress. Many areas of relevant information on negative symptoms and detrimental events were taken into account but potential problems originating from liquidity shocks or default events in the supply chain have been neglected so far in the Early Warning System. The goal of this project is to reduce the corporate credit risk by improving the Early Warning System. This will be achieved by including the impact of potential contagion effects originating from the supply chain on the firm's distress likelihood. This research question is stated as follows: "For each firm what is the probability of integrated default occurring in 3 months?". Customers have integrated default state if they have defaulted or rated close to default due to financial distress.