Application Of Gradient Boosting Model In Bank Product Forecast

In recent years,data mining and cloud computing have been greatly developed,all walks of life have collected a large number of user information.In order to better meet the needs of users,user centric behavior analysis has become the focus of the industry,providing a good user experience has become the key to the development of enterprises.Europe's largest retail banking business of Bank of Santander hope that through the analysis of user behavior prediction,users in the future will increase the purchase of bank products,in order to adjust their business according to the needs of users.This paper uses the public data kaggle platform competition set provided by the Bank of Santander,according to the data recorded before the 17 months to buy the product users,users buy products increase month forecast situation.The original problem is transformed into a multi classification problem.Predict the probability of the user to buy all kinds of products,select the maximum probability of 7 products are recommended to the user in an orderly manner.The whole paper is divided into five steps.The first step is to analyze the present situation and significance of the research.The second step is the analysis and research of the model,which is the theoretical basis for the following experiments.The third step of data preprocessing,exploratory data analysis and data visualization,the visualization analysis can be more intuitive understanding of the meaning of data contains complex data,provide theoretical basis for the engineering characteristics.The fourth step is to deal with the data of the feature engineering,feature engineering plays an important role in the whole data mining process,and the model is the best way to achieve it.Through the extraction,selection and construction of data features,the process is iterated until the model has good results.The fifth step is to optimize the parameters of the model,and get the final results.The results show that the XGBoost model and the LightGBM model are compared with other models in different feature engineering.Product prediction based on gradient boosting model is not only applicable to the Bank of Santander,also this method can be used in various bank products in the domestic business,improve work efficiency and service level of the bank.