Support Vector Machine And Credit Evaluation Model For Individual

In recent years,the credit consumption has been developed very quickly in our country,thanks to the growth of internet finance.And it's very important to evaluate the customers' credit due to the low limitation and high risk of this business.To evaluate customers' credit accurately,the organization have to establish a credit evaluation model,and they also have to decrease the costs that caused by judging the customers' credit wrongly.However,most models only considered the problem that different types of the wrong judgment on customers' category would cause different loss,but didn't consider that different types of the wrong judgment on customers' credit amount could also cause different loss.So,in this paper,we establish a cost-sensitive credit evaluation model base on the classification of the customers' credit amount,to decrease the total loss of the wrong judgment in the process of credit evaluation.Firstly,we divide the samples by the their credit amount,and then divide every set into two subsets by its category.We compare the loss of different types of the wrong judgment on customers' credit amount and the loss of different types of the wrong judgment on customers' category,to get the loss ratio of the every different subsets.Secondly,we build the loss function based on support vector machine,and give the different weights to the different subsets in the loss function,and establish a new credit evaluation model based on cost-sensitive support vector machine.Finally,we solve the optimization problem and implement the algorithm to evaluate customers' credit.We compare the effects of the new model and the traditional support vector machine model by using a real credit data,and the result suggests that our new model can be costsensitive to the category and the credit amount.Moreover,it can decrease the total loss and increase the accuracy of judgment of big-amount customers when used to evaluate the customers' credit.