Research On CRM Of Commercial Bank Based On Machine Learning

With the innovation of information technology and the arrival of the era of big data,data analysis has become the basic means for the development of any industry,and banking industry will also follow the developing trend of times actively.As we all know,establishing stable long-term relationship with customers is the foundation for banks to survive.In order to manage customers more efficiently,commercial banks have introduced customer relationship management(CRM)system.However,the application of CRM by banks has been at the basic stage of confirming expired business and inquiring customer information for a very long time.Faced with the severe competition situation,commercial banks urgently need to explore how to introduce machine learning technology into CRM data mining creatively.The data set formed by customer information in CRM system can be used to predict sales results,formulate marketing strategies matching customer expectations,save bank costs and improve profitability.Firstly,this paper defines the meaning of CRM in commercial banks,establishes the organizational structure of CRM in commercial banks,and recognizes that machine leaming can indeed affect customer life cycle from the analysis level CRM and the operation level CRM.Then,the case of predicting whether customers will buy financial products of commercial banks is taken as an example to start the research of CRM based on machine learning.The experimental platform sklearn is an open source machine learning library based on Python.In this paper,we use Gradient Tree Boosting(GRT)algorithm in the integrated learning method of sklearner library,and use 41]88 data in UCI to train.After learning,the default parameter set training classifier is used to test the test samples.The test results are Accuracy 0.900 and AUC value 0.945.After adjusting the relationship between learning rate and estimator in the way of grid tracing-exhaustive grid search,tuning the number of parameters and reducing the number of learning rates and increasing the number of estimators.The accuracy was raised to 0.911 and AUC to 0.957.In order to further test the accuracy of the trained data for the prediction of financial products of commercial banks in China.We take Shanghai Pudong Development Bank Tiantianyeng Zengli No.1 financial product as an example to test.The prediction accuracy of the test results was 0.900,and the AUC was increased to 0.938.The results show that although the predictive ability is slightly lower than that of data acquisition countries,the significance of machine learning to CRM is affirmed.Based on this model,the relationship between CRM data prediction and the marketing quantity of a product is further studied.On the one hand,data prediction can effectively reduce the number of marketing and reduce the cost of marketing.Taking the test data set of this paper as the object,and aiming at getting 60%of potential customers,the marketing quantity can be reduced from 60.0%of total customers to 14.53%by data prediction.On the other hand,by accurately grasping the expected marketing quantity,the machine learning classifier can be adapted to different marketing characteristics to achieve better results.