Exploring the Impact of Macroeconomic Factors on Credit Default Prediction Using Machine Learning and Neural Network Methods
DOI:
https://doi.org/10.61173/a19rh053Keywords:
P2P lending, Credit default prediction, Machine learning, MacroeconomicsAbstract
Macroeconomics has a profound impact on the financial condition and loan behavior of individual borrowers. This article provides a specific analysis of the macroeconomic impact based on data from P2P lending platform LendingClub. Firstly, traditional machine learning models are used to train predictive data. By comparing the predictive performance of macro sensitive and nonsensitive micro features at similar scales, it is found that macro-sensitive micro features are more suitable for model training; Further inclusion of representative macro indicators leads to an improvement in performance, confirming the direct impact of macroeconomic factors on credit default predictions. Given the limitations of traditional machine learning models in dealing with feature multicollinearity and nonlinear relationship of the data, DNN (Deep Neural Network) models are introduced to optimize the performance. The research results indicate that at the feature selection level, macro-related features have a significant impact on credit default prediction. At the level of model evaluation, neural network models have better learning and default prediction performance compared to traditional machine learning models for imbalanced and nonlinear credit datasets in the real world.
References
To address this, it is necessary to strengthen the nonlinear N., Jumanto, N., Dasril, Y., & Iswanto, N. (2023). New model structure of the network (increasing depth width, replacing combination meta-learner to improve accuracy prediction P2P dynamic attention, adding skip connections), enhance the lending with stacking ensemble learning. Intelligent Systems nonlinearity of input features (nonlinear transformation, With Applications, 18, 200204.
constructing interactive features), optimize training strat- [2] Baltazar, J., Reis, J., & Amorim, M. (2020). Sustainable egies (extending early stopping patience, using learning economies: Using a macro-economic model to predict how rate scheduling, retaining extreme samples), so that DNN the default rate is affected under economic stress scenarios. can fully leverage its nonlinear advantages and capture Sustainable Futures, 2, 100011. complex correlation patterns in data. [3] Zhang Liying, Yang Ruojin. Application Research on Machine Learning-Based Personal Loan Default Prediction 5. Conclusion Models[J]. Financial Regulatory Research, 2022, (06): 46-59. DOI: 10.13490/j.cnki.frr.2022.06.002 Through a series of data filtering and optimization mea- [4] Cao Yujia. Research on the Causes of Personal Credit sures of model and method mentioned above, this study Customer Default and Prediction Schemes in Commercial believes that macroeconomics has a profound impact on
Banks[J]. Finance and Economics, 2022, (26): 51-53. DOI: loan default prediction, which is reflected not only in the use of direct macroeconomic indicators as training data, 10.19887/j.cnki.cn11-4098/f.2022.26.007
but also in the use of macro sensitive micro indicators as [5] Kawano, K., Kutsuna, T., & Sano, K. (2025). Minimal part of the datasets. sufficient views: A DNN model making predictions with more In addition, in response to the three core issues with the evidence has higher accuracy. Neural Networks, 190, 107610.
Lendingclub datasets and corresponding features: mul- [6] Eggertsson, G. B., & Krugman, P. (2012). Debt, ticollinearity between features, non-linear relationships Deleveraging, and the liquidity trap: A Fisher-Minsky-Koo between data, and imbalanced data samples, this study approach*. The Quarterly Journal of Economics, 127(3), 1469– continuously reduces the impact of these issues on the 1513.
model through feature screening, model optimization, and [7] Avery, R. B., Calem, P. S., & Canner, G. B. (2004). Credit targeted method implementation. At the same time, a more report accuracy and access to credit. Federal Reserve Bulletin, realistic evaluation method, WACC, was used to assess 90(3), 0.
the performance of various models, and it is ultimately [8] Stiglitz, J. E., & Weiss, A. (1981). Credit Rationing in concluded that the DNN model is the optimal choice for Markets with Imperfect Information. American Economic predicting loan defaults in this type of scenario. Review, 71(3), 393–410.
However, there is still room for improvement in solving [9] Ganong, P., & Noel, P. (2022). Why do Borrowers Default nonlinear problems, and in the future, the study will con- on Mortgages? The Quarterly Journal of Economics, 138(2), tinue to focus on better utilizing neural network models 1001–1065.
and more appropriate parameter settings to solve this [10] Duan, J. (2019). Financial system modeling using deep problem. neural networks (DNNs) for effective risk assessment and From a practical perspective, this study has practical valprediction. Journal of the Franklin Institute, 356(8), 4716–4731. ue. For financial institutions, the “macro+micro” dual di- [11] Bayraci, S., & Susuz, O. (2019). A Deep Neural Network mensional data fusion approach, targeted data governance (DNN) based classification model in application to loan default plan, and optimized DNN model provided in this study prediction. DOAJ (DOAJ: Directory of Open Access Journals). can help institutions more accurately capture the linkage risk between macroeconomic fluctuations and micro bor- [12] Zandi, S., Korangi, K., Óskarsdóttir, M., Mues, C., &
rower characteristics. The method effectively reduces the Bravo, C. (2024). Attention-based dynamic multilayer graph probability of misjudgment in credit approval, reduces neural networks for loan default prediction. European Journal of default losses, and optimizes the allocation efficiency of Operational Research. credit resources towards low-risk and high return areas, [13] Di, S., Wang, Y., Yang, D., Liu, Y., Zhang, J., & Zheng, W.
thereby improving the sustainability of credit business. (2025). SMOTE-enhanced XGBoost for rapid seismic damage For the borrower group, a more scientific risk assessment assessment of bridge portfolios. Soil Dynamics and Earthquake system can avoid “misjudgments” caused by the limita- Engineering, 199, 109712. tions of traditional models, allowing borrowers with good
Downloads
Published
Issue
Section
License
Copyright (c) 2026 by the authors.
This work is licensed under a Creative Commons Attribution 4.0 International License.
