Loan Default Prediction Based on Convolutional Neural Network and LightGBM

Abstract

With the change of people's consumption mode, credit consumption has gradually become a new consumption trend. Frequent loan defaults give default prediction more and more attention. This paper proposes a new comprehensive prediction method of loan default. This method combines convolutional neural network and LightGBM algorithm to establish a prediction model. Firstly, the excellent feature extraction ability of 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 LightGBM algorithm are adjusted through grid search so as to build the LightGBM model. Finally, the LightGBM model is trained based on the new feature matrix, and the CNN-LightGBM loan default prediction model is obtained. To verify the effectiveness and superiority of our model, a series of experiments were conducted to compare the proposed prediction model with four classical models. The results show that CNN-LightGBM model is superior to other models in all evaluation indexes.

Tab-Attention: Self-Attention-Based Stacked Generalization for Imbalanced Credit Default Prediction

“…Chen et al [7] demonstrated that GBDT outperforms other methods in tabular data applications. Several GBDT algorithms, such as XGBoost [6,14], LightGBM [29], and CatBoost [23], have been developed to predict credit risk successfully. Although these ensemble algorithms exhibit differences, their performance is often similar across various tasks [23].…”

“…In general, existing works have shown that GBDT [6,7,14,23,29], deep learning [13,18,27,28], effective feature selection strategies [3,21], and multi-view ensemble [25,27] could significantly improve model performance. However, there is a limited exploration on how to combine these optimization strategies to obtain better solutions.…”