A Logistic-BP Classifier for Assessing Personal Credit Risk

Li-Qian WU


In view of the present personal credit risk problems in peer-to-peer industry, this paper proposes the Logistic-BP (Back Propagation) combined optimal model. The method combines the traditional statistical method with the artificial intelligence method by building a unified error function. The simulation is carried out with a German commercial banks’ credit data, and the result shows that the Logistic-BP model has higher accuracy and robustness than single model. The average classification accuracy for the test sample is 77.3%, 2-7% higher than the single model and has an obvious promotion effect. In this paper, we compare Logistic-BP with other methods such as LDA, Naive Bayes, RBF-LS-SVM and C4.5, the result shows that Logistic-BP is superior to other algorithms.


Logistic Regression, BP Neural Network, Logistic-BP, Personal Credit Assessment


Full Text:



  • There are currently no refbacks.