Related Experiment Video
Updated: Jul 7, 2026
08:59
Morphology-Based Distinction Between Healthy and Pathological Cells Utilizing Fourier Transforms and Self-Organizing Maps
Published on: October 28, 2018
Bankruptcy analysis with self-organizing maps in learning metrics
S Kaski1, J Sinkkonen, J Peltonen
1Neural Networks Research Centre, Helsinki University of Technology, Espoo, Finland.
IEEE Transactions on Neural Networks
|February 6, 2008
Summary
This study introduces a novel metric for analyzing financial statements, enhancing bankruptcy prediction. The new method uses a self-organizing map (SOM) to highlight key bankruptcy risk indicators within enterprise data.
Area of Science:
- Computational finance
- Data mining
- Machine learning
Background:
- Traditional financial analysis often overlooks local data variations.
- Predicting enterprise bankruptcy requires sensitive metrics that capture subtle data shifts.
Purpose of the Study:
- To develop a novel data-driven metric for exploring financial statements.
- To enhance the visualization and analysis of enterprise financial data for bankruptcy prediction.
Main Methods:
- Deriving a local metric based on the Fisher information matrix.
- Applying a self-organizing map (SOM) in the new metric space.
- Estimating conditional densities of an auxiliary variable indicating future bankruptcy.
Main Results:
- The new metric effectively measures local distances based on changes in bankruptcy indicators.
- The SOM, using the new metric, preserves data topology while highlighting directions of highest bankruptcy probability change.
- This approach improves the visualization of financial data concerning bankruptcy risk.
Conclusions:
- The proposed Fisher information matrix-based metric offers a powerful tool for financial statement analysis.
- This method enhances the interpretability of enterprise financial data for early bankruptcy detection.
- Self-organizing maps in this metric space provide topology-preserving visualizations sensitive to bankruptcy risk.
