One of the distinguishing features of Valuatum Platform is the utilization of state-of-the-art bankruptcy risk models in credit risk estimation. Valuatum has for a long time utilized complex algorithms in evaluation of companies’ bankruptcy risk. Historically this has been done using models based on logistic regression but with emergence of more powerful machine-learning algorithms we have been able to take the bankruptcy risk estimation to another level.
Below we briefly describe different bankruptcy risk models we are currently using. While machine-learning models are the cornerstone of bankruptcy risk estimation in Valuatum system, more simpler logistic regression models can still be useful as they are easy to understand and visualize.
While bankruptcy risk models based on logistic regression are relatively easy to understand and are good for visualizing the bankruptcy risk, they have the downfall of using constant weights for variables that are included in the model. For example, while in general bad liquidity can be alarming when evaluating the credit risk, for highly profitable and financially sound companies liquidity can be a rather irrelevant factor, because they would be able to improve their liquidity at any time. Logistic regression models are bad at catching relationships like this. That is why we have invested countless of hours into the development of state-of-the-art machine-learning models for bankruptcy risk estimation. Compared to regression models, machine learning models are much more versatile and can observe more complex and dynamic relationships between different variables.
- We have studied several different machine-learning algorithms and are constantly looking for ways to improve the model by tweaking the parameters and including additional data points.
- Currently we utilize the Gradient boosting model using XGBoost library, which was used to obtain the best results in testing. Other models we studied include neural networks and random forest model.
- Our bankruptcy risk model is trained with dozens of variables from nearly 200 000 Finnish companies. The current version of the model uses 30 explanatory variables, including industry-specific bankruptcy risk.
- Based on the bankruptcy risk we can build classification models that help our customers to determine the creditworthiness of any company, credit rating, or even the suitable interest rate. All of this information, alongside key financial information, can also be included on customizable, automatically generated reports that can be printed from the system with a click of a button.
- Due to economic cycle and other external factors the bankruptcy risk percentage alone may not always be the best indicator of risk. Therefore, we also show graphically and numerically any company’s position relative to all other companies for any given year.
- Bankruptcy risk is further broken into pieces showing the most important factors that contribute to the risk value.
More information about machine-learning models in Valuatum system
Bankruptcy risk estimation of companies is automatically in the Valuatum Platform using the methods mentioned above. You can read more about the different methods, their technical details, model training and validation, and the way they are implemented in our system for bankruptcy risk estimation in this document Automated estimation of bankruptcy risk.
- The system also calculates the bankruptcy risk estimates for industries based on Finnish default and bankruptcy data and financial statements data
- This provides a good benchmark for any company, as it provides a way to compare the company’s bankruptcy risk to the risks of similar companies in the same industry
- The industry risk level is also included as an explanatory variable in our bankruptcy risk models
- The industries are defined using the NACE standard, which is the standard classification system of industries in the European Union
Logistic regression models
Single variable models
- The picture on the right illustrates the connection between profitability (ROA %) and default risk – for example, as the picture shows, bankruptcy or default risks of companies with positive ROA % stay rather low, but right below the zero, the risk function gets steeper.
- Risk estimations are based on Finnish default and bankruptcy data and financial statements data of nearly 200 000 companies.
- Companies are grouped to large enough groups to make the risk estimates statistically significant.
- However, like the two variable model below strongly shows, the one variable model gives only a limited view on bankcruptcy risk, and that is why Valuatum system also uses models with two or more variables.
Multiple variable models
- Multiple variable model gives more accurate bankruptcy and payment default risk estimations than single variable models
- Any key figures and ratios can be used in the model
- Development of the risk of the examined company can also be illustrated in the graph
- For example, (see the graph on the right), bankruptcy risk of companies with the same profitabilty (ROA %) level can actually vary a lot depending on their solvency (Equity ratio), and even companies with poor profitability can have low bankcruptcy risk if they have high solvency.
Risk by components
- Valuatum system estimates overall bankcruptcy and default risks using several selected components to ensure best possible accuracy
- Overall statistical backruptcy risk can be divided to components
- Easy to interpret graph illustrates which factors have an influence on the bankruptcy risk of the company, and shows how strong the influence is compared to other factors