Feature Selection On Customer-churn Model In Broker

In today's information age,data acquisition is becoming more easier.In the field of data mining,the amount of data that needs to be faced is also larger.In the two-category task,we often encounter data with dimensions up to hundreds of thousands.To use this data to build an effective classification learner,feature selection is an inevitable problem.Before selecting the feature selection method,it is necessary to preprocess the data and process the missing values in the data.When the categories are not balanced,methods need to be taken to reduce the imbalance.There are many methods for feature selection,but there is no unified method for each problem.This paper introduces several methods,and uses these methods to conduct experiments on the brokerage customer churn model,comparing the advantages of different methods.Inferior,choose the method that applies to the existing problem.