Trade risk identification and prediction using experts’ knowledge and machine learning

This study represents a framework of expert knowledge incorporation into the machine learning pipeline used for the integrated trade risk assessment process. We review the related literature to reflect the newest risk prediction methodologies and combinations of multiple approaches, i.e., expert systems, ensemble models application. We integrate expert risk knowledge into the clustering results review process and adjust results based on the risk expert recommendations. Classes confirmed by the expert are further used as labels for the automatic risk assessment process. The framework was tested with the historical data from the trading company, and classification results were evaluated by calculating the F1 score.