Research On Loan Default Prediction Based On Convolutional Neural Network And Ensemble Learning

Since the reform and opening up,especially after China’s accession to the WTO,residents’ consumption habits have undergone tremendous changes,and credit consumption has become a new consumption trend.In order to promote household consumption and boost economic development,the state actively supports the development of credit and other financial services.In recent years,in the context of increasingly fierce competition in the financial market,financial institutions such as credit companies have begun to provide various types of financial products to the public to increase income.However,behind the large amount of financial borrowing lies the huge financial risk of debt default,which may lead to huge economic losses if these institutions fail to monitor and warn of risks in a timely manner.Therefore,the establishment of a scientific and effective risk assessment model is crucial for the healthy development of the credit industry.As a result,loan default prediction has become one of the focus issues of financial and academic circles.Aiming at the problems of complex financial loan data structure and low prediction accuracy of traditional models,this paper proposes a new comprehensive loan default prediction method.This paper combines convolutional neural networks and machine learning integration algorithms to build a hybrid prediction model.First,the preprocessing of the original dataset is carried out,including missing value filling,outlier removal,etc.In addition,feature engineering is a necessary preparation work before modeling,in order to reduce data complexity and improve computational efficiency,this paper introduces a variety of feature screening techniques in feature engineering to ensure the validity of the results.Aiming at the problem of data imbalance,this paper uses synthetic minority oversampling technology(SMOTE)to balance the data to improve the prediction effect of a small number of defaulting customers.At the same time,the excellent feature extraction ability of the convolutional neural network is used to extract features from the original loan data and generate a new feature matrix.Secondly,the new feature matrix is used as input data,and the parameters of algorithms such as Logistic Regression,Random Forest,XGBoost,Light GBM are adjusted through grid search,so as to establish each machine learning model and generate an ensemble model.Finally,an ensemble model is trained based on the new feature matrix to obtain a CNN-LRXL loan default prediction model.In order to verify the effectiveness and superiority of our model,this paper conducts a series of experiments to compare and analyze the proposed predictive model with four classical models and the combined model of neural network and single machine learning.The results show that the CNN-LRXL model is better than other models in terms of ACC,AUC and other evaluation indicators.