Fig. 21
From: A dynamic credit risk assessment model with data mining techniques: evidence from Iranian banks
The optimal threshold. The performance of the model is evaluated by two indicators: 1) Degree of sensitivity: A proportion of bad customers that the model classifies in a bad customer group. 2) The degree of diagnosis: A proportion of well-off customers that the model classifies in a good customer group. In order to judge customers and group them into low risk, medium risk, and high-risk customers, the probability of default within the interval [0, 1] was determined. If the probability of defaults were on more than one point, the customer was considered with a high risk and less than that means the customer is low risk. In this paper, the optimal threshold limit equals a value that maximizes the degree of sensitivity and the degree of diagnosis of the model. This Fig shows that the optimal threshold (Y) of the degree of sensitivity and degree of detection is 0.37. The blue line is the degree of sensitivity and the red line is the degree of diagnosis