Research On The Establishment And Application Of The Label Of Churning Customers In Banks

A bank is a special financial institution.The reason why it is called special is that it is not only a place where money can be stored and traded,but also an institution that has various functions and is closely related to the national economy.In recent years,the competition among commercial banks has gradually escalated,and the emergence of various new payment methods has brought external pressure to banks.The loss of bank customers has become prominent,and there are also many difficulties in developing new customers.How to retain existing customers has become a key point of concern.In the exploration of new business attempts and new service models,banks are constantly trying to combine business and technology to tap the potential behind information.User portrait is one of the important research directions.This article uses customer data from a bank to establish a customer profile system from the perspective of customer churn,and conducts customer research on churn issues.The main work done in this article includes:(1)Through the statistical analysis of the original data,to understand the structure and characteristics of the original data,it is found that the sample of lost customers is relatively small,and there is a serious imbalance in the data,which will affect the analysis results.In order to allow the algorithm to process data better,this article uses one-hot encoding to process classified data,and MIN-MAX standardization to process numerical data.After the data preprocessing is completed,the SMOTE algorithm is used to deal with the imbalance of the data.(2)Seven kinds of classification models were established to predict customer churn.In order to prevent the model from overfitting,ten-fold cross-validation is used to divide the sample training set and test set.In order to further optimize the model structure,a grid search method is used to select model parameters.After comparing the scores of various evaluation indicators of the model,XGBoost with the best performance is selected as the final churn prediction model.(3)In the past,most scholars only made predictions for the problem of churn.This article will further analyze the churn customer characteristics from the perspective of portraits.Therefore,finally,according to the prediction results,attribute labels and rule labels were established from the perspective of customer attributes and bank business.At the same time,referring to the classic customer analysis model RFM,six types were established through principal component analysis and K-Means clustering algorithm.The abstract labels are important keeping customers,important retaining customers,important developing customers,general keeping customers,general retaining customers and general development customers.This type of label summarizes the characteristics of customers in terms of value,risk,and so on.It is simpler and easier to understand than traditional user labels.From the data analysis in this article,we can see that the bank's important keeping customers is more,indicating that most of the bank's customers with high value and high potential are on the verge of churn,which will inevitably affect the bank's future development.Therefore,the relevant staff of the bank should provide accurate services to these customers based on their characteristics in order to achieve the goal of retaining customers and maximizing benefits.