Multimodal Generative Models for Bankruptcy Prediction Using Textual Data

TL;DR

The Conditional Multimodal Discriminative (CMMD) model is introduced that learns multimodal representations that embed information from accounting, market, and textual data modalities that are used to make bankruptcy predictions and to generate words from the missing MDA modality.

Abstract

Textual data from financial filings, e.g., the Management's Discussion & Analysis (MDA) section in Form 10-K, has been used to improve the prediction accuracy of bankruptcy models. In practice, however, we cannot obtain the MDA section for all public companies, which limits the use of MDA data in traditional bankruptcy models, as they need complete data to make predictions. The two main reasons for the lack of MDA are: (i) not all companies are obliged to submit the MDA and (ii) technical problems arise when crawling and scrapping the MDA section. To solve this limitation, this research introduces the Conditional Multimodal Discriminative (CMMD) model that learns multimodal representations that embed information from accounting, market, and textual data modalities. The CMMD model needs a sample with all data modalities for model training. At test time, the CMMD model only needs access to accounting and market modalities to generate multimodal representations, which are further used to make bankruptcy predictions and to generate words from the missing MDA modality. With this novel methodology, it is realistic to use textual data in bankruptcy prediction models, since accounting and market data are available for all companies, unlike textual data. The empirical results of this research show that if financial regulators, or investors, were to use traditional models using MDA data, they would only be able to make predictions for 60% of the companies. Furthermore, the classification performance of our proposed methodology is superior to that of a large number of traditional classifier models, taking into account all the companies in our sample.

Figures (11)

• Figure 1: A graphical framework for multimodal learning in which we have access to 3 data modalities: handwriting, images, and text, describing the same object, a digit. Multimodal learning models relate these sources of information to learn a data representation that embeds information on all data modalities.
• Figure 2: Architecture of a neural network that learns the parameters $\bm{\mu}$ and $\bm{\sigma}$ of a Gaussian distribution. The output layer $\bm{\mu}$ has the same number of neurons as the dimensionality of ${\bm{z}}$. Then we draw a representation ${\bm{z}}$ by using the location-scale transformation ${\bm{z}} = \bm{\mu} + \bm{\sigma} \odot \bm{\epsilon}$, where $\bm{\epsilon} \sim \mathcal{N}(\bm{0},\bm{1})$ and $\odot$ is the element-wise product.
• Figure 3: Architecture for the CMMD model that is composed by 4 neural networks: encoder, decoder, prior, and classifier. The dotted arrow indicates a forward pass during training, which is replaced by the dashed arrow at test time, i.e., the input to $p(\bm{x}_{\mathcal{M}}|\bm{x}_{\mathcal{O}},{\bm{z}})$ is ${\bm{z}} \sim q({\bm{z}}|\bm{x}_{\mathcal{O}},\bm{x}_{\mathcal{M}},y)$ during training, while ${\bm{z}} \sim p({\bm{z}}|\bm{x}_{\mathcal{O}})$ at test time. The solid arrow depicts a common forward propagation during training and testing, i.e., the input to $p(y|{\bm{z}})$ is always ${\bm{z}} \sim p({\bm{z}}|\bm{x}_{\mathcal{O}})$.
• Figure 4: (a): We collect 31 accounting and 2 market predictors that are merged with an MDA from the same quarter. We roll-over the MDA data for 3 more quarters and merge it with the corresponding predictors in those 3 quarters, i.e., an MDA can be used during 4 quarters. (b): The data set in this research includes bankruptcies from 1994 to 2020. We use data from 1994 to 2016 for model training, while data from 2017 to 2020 are used for testing the predictive power of the models. For 1-year forecasts, for example, the latest quarter in which we can make a forecast is 2019Q4, which corresponds to bankruptcies during 2020Q4.
• Figure 5: Number of yearly bankruptcies and bankruptcy rates in our data set for 1-year bankruptcy predictions. Bankruptcy rate includes firm-quarter observations for which we could either merge the 3 data modalities or merge the 3 data modalities with 1-year ahead bankruptcies, while the number of bankruptcies include all listed companies that file for Chapter 7 or 11.