Bankruptcy prediction using imaged financial ratios and convolutional neural networks

Highlights：

• Convolution networks can predict bankruptcy by inputting financial ratios as an image.

• Predictive accuracy improves with correlated financial ratios placed in the vicinity.

• Deeper network configuration improves predictive accuracy.

• Creating artificial financial data does not ensure the same effect as using real data.

• Convolution-network-based bankruptcy prediction outperforms traditional methods.