On Prediction Of Deposit Clients Based On Classification Ensembles

Nowadays,due to the rapid development of Internet,large amounts of Internet financial derivatives have been emerged.With the changing consumption concept of people,it results in a nagative impact on the deposits that is the root of the banks.More and more people invest some of their deposits into new financial products or consumption.Meanwhile the banks have a large amount of customer data,it can be used,based on data mining technology,to identify customers with deposits willing and thereafter improve marketing efficiency of the banksThis paper focuses on the combination of data mining technology and prac-tical business to guide the marketing activities of Banks.On the one hand,through the descriptive analysis of data characteristics,the marketing strategy of Banks is given from the qualitative perspective,and the new features are constructed by using bayesian smoothing method and business characteristics.On the other hand,the model of decision tree,random forests,xgboost and LightGBM are used to select the optimal parameters to obtain the optimal model.Combined with the unbalanced data processing method,the above optimal model is improved to get a more effective classifier.By comparing the model effects,it is found that the model with newly constructed features and processed data imbalance has a higher degree of clients recognition.In the end,the number of clients to be marketed was reduced to one sixth of the origin.These clients included 82.4%of the origin who wanted to deposit in the bank.The marketing strategy guidance from the analysis and the final model can effectively reduce the cost of the bank and carry out targeted marketing to customers,which can effectively achieve the purpose of increasing the bank's deposit.