Research On Customer Credit Evaluation Model Based On Big Data Method

With the increasingly fierce international competition,the increasing pressure of competition in the domestic market,the continuous development of China's economy and the deepening of the reform of the market economy system,enterprises have gradually established their own customer credit evaluations system in order to expand the scale of product sales and increase the market share of products.However,due to the lack of objectivity,dynamics and integrity of the traditional corporate customer credit evaluation model,the company cannot fully and timely understand the credit status of the evaluation object.At the same time,with the advent of the Internet and the era of big data,the evaluated companies exposed more credit-related information to the public's field of vision,such as related parties.Therefore,it is necessary to use big data technology to incorporate more useful information into the credit evaluation system,and to find out the hidden logical relationship between various information,and to study the repayment ability and default risk of enterprises more efficiently and quickly.First of all,this thesis sorts out the problem of under-utilization of big data technology in traditional customer credit evaluation mode,and sorts out the customer credit evaluation process based on big data method.Then,combined with the characteristics of more information exposure in the era of big data,the thesis were selected preliminarily indexes combined with big data technology through the sorting of relevant literature and the author's investigation.,and then the author used the grey correlation clustering analysis method to screen the evaluation indicators again,eliminating the redundant indicators and determining the final customer credit evaluation index system.Then,in order to ensure business interpretability,model stability and model accuracy,this author constructs a weighted fusion credit evaluation model for logistic regression and gradient lifting decision trees.In addition,the thesis also uses the ten-fold cross-validation method to test the accuracy and rationality of the model,and compares the Z-score model method,logistic regression model,gradient-lifting decision tree model,the weighted fusion model of logistic regression with gradient-lifting decision tree,The result shows that the weighted fusion model has a higher accuracy rate of more than 75%.Finally the thesis compares the evaluation results of the model based on big data method with the customer historical credit data of Z Company.The results prove that the customer credit evaluation model based on big data method has 50% accuracy and the orher 50% sample error range is within [-0.08,0.08].so it also fully proves that the customer credit evaluation model based on big data method has higher accuracy and feasibility.