Financial Ratios, Deep Learning and the Prediction of Corporate Credit Events
碩士 === 靜宜大學 === 財務與計算數學系 === 105 === The traditional models assume linear relationship on the explanatory variables but, in the real economic world, economical factors have non-linear relationship. Thus linear assumption on variables limit the prediction accuracy. This study improves the accuracy of...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | zh-TW |
| Published: |
2017
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/pgzmt9 |
| Summary: | 碩士 === 靜宜大學 === 財務與計算數學系 === 105 === The traditional models assume linear relationship on the explanatory variables but, in the real economic world, economical factors have non-linear relationship. Thus linear assumption on variables limit the prediction accuracy. This study improves the accuracy of the default model with linear assumption by using nonlinear properties in deep learning. Deep learning models introduce multi-linear transformation and add non-linear relationship on the factors to classify the default events. Using deep learning on financial ratios, default models is built for Taiwan’s exchange and OTC market that tries to improve the prediction accuracy of default events.
|
|---|