Design And Implementation Of Early Warning System Of Bank Customer Churn Based On Customer Portrait

With the advent of the era of big data,competition in the financial sector has intensified,and commercial banks have become increasingly concerned about customer churn management.At the same time,commercial banks use their accumulated massive customer data to create customer portraits to depict customer characteristics.However,traditional financial customer portraits only show basic customer information and some statistical data,ignoring the hidden information behind the data.In this thesis,the retail customers of a commercial bank are taken as the research object,and the problem of customer churn is studied in depth.It is found that not all the churned customers are worthy of retention by the bank,and attention should be paid to the churn of highvalue customers.At the same time,customer value clustering provides value judgment for loss prediction,which is an effective way to improve customer loss prediction and retention ability Therefore,an early warning system of bank customer churn based on customer portrait is designed and developed in this thesis,collecting and processing customer related data in the customer portrait platform,applying customer value clustering model and customer churn prediction model,so as to give early warning in time before customer churn,retain high-value customers and reduce customer churn rate.The main work of this thesis is as follows:(1)A customer value clustering model based on customer portrait is proposed.First of all,aiming at customer value measurement,an evaluation index system of bank customer value based on RFM model and CV-PV model is proposed.Secondly,in the cluster analysis of customer value,a K-Means algorithm based on maximum and minimum distance and weighted density is designed.The weighted density is used to select the cluster center set,and the initial cluster center is selected according to the maximum and minimum distance.The experimental results show that the clustering effect of this algorithm is better than that of the basic k-means algorithm.(2)A customer churn prediction model is constructed based on the customer portrait.First of all,in view of the class imbalance phenomenon of bank customer churn data set,the cost sensitive learning method is integrated into the random forest algorithm,and the customer churn prediction model is constructed.At the same time,according to the actual distribution of data and the importance of features,the weight distance is introduced to reconstruct the cost function,and the weight voting is carried out according to the performance of the basic classifier to improve the accuracy and overall effect of classification.Secondly,the customer churn prediction model is used to predict the churn of the total customer and the customer group after segmentation respectively.The comparison shows that the prediction effect of customer value segmentation is improved compared with the total customer indicators,And the accuracy of core value customer churn prediction increased by 6.67%,AUC value increased by 4.5%.Finally,by comparing with other algorithms,it is found that the weighted random forest algorithm based on cost-sensitive has better performance in predicting bank customer churn.(3)An early warning system of bank customer churn based on customer portrait is designed and implemented.The system consists of data acquisition module,data modeling module and data visualization module.The functional and non-functional tests are carried out to verify the effectiveness and feasibility of the system.