Fig. 7: Credit scoring by CIM.
a The score distributions for normal and overdue repayments. The non-overlapping area between the two distributions indicates the effectiveness of the feature selection process. b Definition of KS (Kolmogorov–Smirnov) statistic, which measures the maximum difference between the cumulative distributions of different labels. c As the alpha parameter increases, the KS statistic of the quantum feature selection CIM + XGBoost model rises. When alpha exceeds 0.99, the quantum feature selection CIM + XGBoost model outperforms the XGBoost model without feature selection
