Research On Precision Marketing Strategy Of Banks Based On Data Mining

With the advent of the era of big data,data mining and data analysis technologies have been widely used in various fields.The financial field has always used the latest technology to continuously update its own management and management methods.As a relatively mature banking industry in the financial industry,the way of selling financial products has been greatly changed from the traditional purposeless form.It uses big data for intelligent analysis,and uses algorithms to build models to make accurate Forecast,carry out diversified and purposeful precision marketing.Facing the huge customer information data set in bank marketing activities,how to select the most suitable algorithm model for prediction is an unavoidable important problem for banks to improve their core competitiveness.This article will model and study how to conduct precise telemarketing for different types of customers.Considering that the banking industry is geared towards people from all walks of life,the collected customer information data sets are seriously unbalanced,and the types of customer related attribute variables are also not the same.Starting from the data level,this article optimizes the data set for the processing of unbalanced data sets,unification of different types of variables,etc,combined with data mining tools,comprehensively considers multiple evaluation indicators,and selects the best fit for the data set.A better model to improve our prediction accuracy of customer propensity to facilitate precision marketing.Finally,according to the predicted results,the customers are classified into different customer groups,and then the user characteristics of different customer groups are mined,and the marketing plan is formulated in a targeted manner to maximize the bank's revenue and reduce the marketing cost as much as possible.Starting from these two aspects,provide practical and effective suggestions for telemarketing strategies.The experimental data set for this subject comes from the open source data website UCI,which is a customer information data set of a Portuguese bank.There are 45,211 items in total.The positive samples account for 11.7%,the negative samples account for 88.3%,and the positive-negative ratio is close to 1:9.After preprocessing the data,we use SMOTE oversampling to sample the training set,and then use logistic regression,random forest,Light GBM,and XGBoost models to train the data.Use multiple evaluation indicators to evaluate each classification model.