Prediction Of Overdue Repayment Probability Of Staged Shopping Users Based On Convolutional Neural Network

With the rise and rapid development of the Internet finance industry,the staged shopping model,that is,the consumption pattern that consumers can pay for multiple times,has been developed and is increasingly favored by consumers.But at the same time,Internet financial institutions are also facing financial risks.Therefore,whether the user can accurately and effectively predict the repayment behavior of the user will help the financial institution to reasonably evaluate the user's credit risk and complete the rational delivery of the funds,so as to ensure the safety of the funds to the utmost extent.At present,the overdue repayment probability prediction mainly uses the traditional machine learning method.This method needs to analyze the data and establish the feature engineering.Finally,the prediction model is built according to the extracted features,but with the increasing data types and the complexity of ascension,the establishment of the characteristics of the project is becoming more difficult,which directly influence the final prediction effect.Convolutional neural network is a multi-layer neural network structure constructed by the visual mechanism of counterfeit organisms.It has been successfully applied in the fields of image processing and speech recognition.The convolutional neural network has the function of automatically extracting high-dimensional features from data and realizing classification prediction,and solves the problem of complex data feature extraction.It can also be applied to the field of overdue repayment probability prediction.Therefore,this paper introduces a convolutional neural network model to predict the user's overdue repayment probability and compare it with the traditional machine learning prediction method.The main research work of this paper is as follows:(1)Combining the user data of the staged shopping website,this paper proposes an overdue repayment probability prediction model based on the improved convolutional neural network,and user data are divided into basic information data and time series data,andsub-convolution neural network is constructed to extract its features,which avoids the problem of large-scale construction of artificial features,saves labor and time costs.Finally,a comparative experiment with the traditional machine learning model was established.The experimental results show that the improved convolutional neural network model is equivalent to the XGBoost model in traditional machine learning under the premise of omitting feature engineering.(2)In order to further improve the prediction accuracy and solve the problem that the convolutional neural network model is difficult to adjust,the improved convolutional neural network model and XGBoost model are fused to form CNN-XGBoost model by using ensemble learning stacking method.The advantages of convolutional neural network in automatically learning high-dimensional features were combined with the powerful classification prediction ability of XGBoost to establish the model and predicts the prediction of overdue repayment probability of shopping users.The CNN-XGBoost model achieved better prediction results by comparing with the existing models.(3)Designing and implementing the overdue repayment probability prediction system for staged shopping users based on the actual user data of the staged shopping website.