Research And Application Of C5.0 Algorithm In Personal Credit Evaluation

This year,Baixing Credit Information Co.,Ltd.applied to the People's Bank of China for personal credit information business and was accepted.This shows that China has officially entered the era of universal credit.For Internet finance companies,how to make reasonable and accurate predictions of personal credit will become a key issue in reducing the risk and operating costs of financial institutions.Based on this background,this paper designs a model for classification and evaluation of personal credit based on C5.0 decision tree algorithm and designs a personal credit evaluation platform based on this model.Firstly,based on the principle of machine learning,this paper compares the difference between decision tree algorithm and common machine learning algorithm,and theoretically proves the improvement process of decision tree algorithm from ID3 algorithm to C4.5 algorithm and then to C5.0 algorithm,demonstrate the superiority of the improved C5.0 algorithm in constructing the personal credit classification model,and the superiority in performance.Secondly,based on the theoretical principle of C5.0 algorithm,the optimal feature selection module,EBP pruning module,Adaboost integration module and cost matrix setting module of ID3,C4.5 and C5.0 algorithms are programmed.Through the real personal credit data set of a bank,the ID3 model,the unpruned C4.5 model,the pruned C4.5 model and the modified C5.0 model were trained to verify the integration and the cost matrix setting optimization achieve the validity to improve C5.0 model.Then through the experimental comparison and analysis,the adjustment process of the improved C5.0 model parameters is achieved,and the optimal personal credit classification model is obtained.The reliability of the improved C5.0 model and the scientific and objective nature of the realization processis is verified by comparison with the experiment under the same conditions in the journal literature.At the same time,the optimization strategy of improving C5.0 model in practical application is proposed.Finally,the personal credit evaluation WEB platform is designed based on the optimal C5.0 model,and the interaction design of the model in practical application is enhanced.Experimental results show that the improved C5.0 personal credit classification model compared with ID3 model,the accuracy rate on the test set increased by 24.41%,and increased by 15.64% compared with C4.5 model.Compared with the common machine learning model,the improved C5.0 model has the highest accuracy on the test set while ensuring the high accuracy on the training set,that is,the improved C5.0 personal credit classification model is more reliable.