Paper: Bayesian Networks Improve Out-of-Distribution Calibration for Agribusiness Delinquency Risk Assessment
Fill the form and get access to the publication
Performing an agricultural credit risk analysis often means making decisions based on incomplete, scattered, and inconsistent data. This is because agribusiness is an economic activity whose risks are extremely difficult to predict. Numerous independent variables interfere with the borrower’s ability to pay: weather conditions, exchange rate variation, quality of inputs, planted culture, among many others.
“And it is in this aspect that the computational capacity of a trained machine is very efficient. It has the ability to extract and review an unimaginable amount of data. This, of course, in a humanly impossible time.
In this scientific study, conducted by the Traive team, you will understand how artificial intelligences need to be trained to handle the complexity involved in a credit risk analysis. The article has passed through the scrutiny of a technical panel from the prestigious Association for Computing Machinery (ACM) and was presented at the International Conference on AI in Finance 2023, in New York.