An Study Of Using The Data Mining Technology To Predict The Success Rate Of Bank Telemarketing

Deposits are the main source of banks while deposits with high liquidity are the guarantee of banks to achieve steady development.Nowadays,with China’s financial development being better,commercial banks face a wide range of competition.Bringing sufficient funds for the bank by a good deposit business will be of the crucial factor to achieve business transformation and improve the competitiveness.Telemarketing is a kind of direct sales,selling products or providing services by phone.Because of its low cost,high efficiency and high degree of freedom,telemarketing began to develop rapidly from the production.In China,especially in the insurance industry,it is widely used.Commercial bank telemarketing is mainly achieved in the customer service center,making full use of its customer information,high trust advantages.However,telemarketing is still blind marketing,customer disgust and other issues,so how to choose the right target customers to improve telemarketing is an important issue facing the marketing manager.This paper mainly uses the data mining technology to conductthe modeling analysis using the bank telemarketing history data and select the target customer.The main contents and conclusions are:(1)Designing of the main process of this study by studying major steps of data mining.Analyzing the bank customer data and the influence of each variable on the marketing result through descriptive statistics.(2)By comparing the effect of standardized data processing and unbalanced data processing on data mining results,it is found that the standard deviation of data and the resampling method can both improve the model prediction effect.(3)Proposed a method which select the characteristic variables by the combination of descriptive statistics and variable sensitivity analysis.This method can simplify the work and eliminate the disturbance variables and improve the model efficienct when the data characteristic variables are constant.Finally,according to the experimental results to eliminate data center customer work variables.(4)Using the decision tree,neural network and support vector machine to build the models on the bank telemarketing data.Through the performance evaluation indexes such as the operating characteristic curve and the LIFT cumulative curve,it is found that these classifiers can effectively improve the classification result and the neural network model is the best.(5)Through the sensitivity analysis,it tests the correlation between all the characteristic variables and the neural network model.By analyzing the contact information group variables,the marketing strategy is provided for the marketing manager,while the marketing activities can be adjusted to the end of each quarter and the talk time is controlled within the humidity range.