Loan risk classification research of commercial bank based on improved SOM network (ID: 2-090)

Loan risk classification has exploited active action in risk administration in commercial bank. The traditional SOM network's weight adjustment is determined only by its learning rate and the difference between the input pattern and the winner neuron's weight. It seems that the SOM obvious ignores some correlation relationships during the learning, which actually exist between the input pattern and the weights of all the nodes that participate in competition, affect the effect of the SOM network pattern classification. This paper presents a BSOM network, find the dependency relation between the nerve cell in the competitive layer and the inputted sample, and select the win node. It acquires better result in experiment that 100 unlisted enterprises'. loan risk classification of a commercial bank.