Search

In order to improve the accuracy and computational efficiency of the credit risk assessment model in an environment of large data, a parallel convolution neural network (CNN) model for credit risk assessment based on Mapreduce framework is proposed. First, the framework of Mapreduce parallel system and its composition module are given; secondly, the convolution neural network is introduced, and the parallel computing model of credit risk assessment is built based on the Mapreduce parallel system framework. Finally, by comparing the experimental results on the credit risk assessment test set of ICBC, the proposed algorithm is shown to have: i) higher assessment accuracy, ii) lower assessment error and iii) less computational time. The effectiveness of the proposed algorithm is thus verified. [ABSTRACT FROM AUTHOR]