Corporate risk stratification through an interpretable autoencoder-based model
提出一种基于自编码器和最近邻密度估计的可解释机器学习模型，将企业资产负债表投影到二维空间，直观识别财务困境区域，帮助非金融企业预警风险。
In this manuscript, we propose an innovative early warning Machine Learning-based model to identify potential threats to financial sustainability for non-financial companies. Unlike most state-of-the-art tools, whose outcomes are often difficult to understand even for experts, our model provides an easily interpretable visualization of balance sheets, projecting each company in a bi-dimensional space according to an autoencoder-based dimensionality reduction matched with a Nearest-Neighbor-based default density estimation. In the resulting space, the distress zones, where the default intensity is high, appear as homogeneous clusters directly identified. Our empirical experiments provide evidence of the interpretability, forecasting ability, and robustness of the bi-dimensional space. • Popular models for corporate risk assessment often provide results that are difficult to understand. • A novel visual-based approach to detect potential distress and safe zones is proposed. • Balance sheet data are mapped into an easily interpretable bi-dimensional space. • An innovative autoencoder-based model is developed to define the reduced space. • The reduced space has high explanatory power and strong predictive ability of the current and future status of the companies.