Research On The Intention Prediction Model Of Bank Customers Purchasing Products

With the rapid development of "Internet +" and the continuous deepening of the application of big data technology,various industries have accumulated a large amount of user behavior information.Based on better meeting user needs,user-centered behavior analysis has become the focus of various industries,and providing good user experience services is the key to enterprise development.Traditional commercial banking has a large volume and a large number of users.The bank expects to analyze the historical behavior of users and predict that users will increase the purchase of bank products in the future,in order to adjust their own business according to the needs of users,and adapt to the trend of Internet finance development.This paper studies and analyzes the historical behavior data of bank users,and proposes a model of intention prediction for bank customers.Firstly,the data is subjected to feature engineering processing,and the input feature vector of this paper is constructed through data cleaning,classification and screening.Secondly,support vector machine(SVM),least squares support vector machine(LS-SVM)and contrast models XGBoost and LightGBM are used to train on the dataset,and the grid traversal method with cross-validation is used to model the parameters.Optimization,parameter adjustment and optimization work for penalty factors and kernel functions,respectively,to find the optimal combination of parameters,get better results of the model.Finally,the performance of the support vector machine model is compared with other models such as XGBoost model and LightGBM model.The empirical study uses the public data set of Santander,Spain.Data statistical analysis technology is used to classify statistics of data characteristics and analyze the implicit information of the original data.Based on the statistical analysis structure of the data,the feature extraction of the data is carried out by feature engineering,and the data features of the paper are constructed from three aspects: initial feature,simple feature and complex feature.Further,the model is used to predict the purchasing tendency of bank customers,and based on the historical behavior records of the products purchased by the user in the first 17 months,the user is predicted to purchase new products in the current month.Bank customers may add more purchases of multiple banking products.This article turns the original problem into a multi-category problem,sorts according to the probability of predicting users to purchase various products,and selects the products with the highest probability to be recommended to users in an orderly manner.Finally,a general evaluation indicator is used to evaluate the performance of the model.The research finds that the optimal combination parameters are obtained in this paper.The best effect is obtained by LS-SVM with linear kernel function,and the classification accuracy rate is 87.41%.This paper proposes that the bank customer purchase product intention prediction model can be applied to the domestic major banks’ product business,thereby improving the bank’s work efficiency and business level.The innovations and contributions of this thesis are as follows:Firstly,a fusion feature extraction framework and method are proposed.The feature engineering is used to clean,classify and reconstruct the original features,and a new feature is formed.The feature extraction method can be better.Effectively express the characteristics of the historical purchase behavior of bank customers.The second is to construct a predictive model of the support vector machine to predict the multi-classification problem of the purchased product.In the process of constructing the optimal parameter model,the grid traversal method is used to optimize the parameters of the support vector machine,and the training is used to obtain the optimal prediction model,which is then used for the customer’s future intention to purchase products.The work of this paper has important reference value for bank customers’ intention prediction research of future purchase products,and provides decision-making basis for banks to recommend wealth management products to customers.