Early Warning of Financial Distress Using Clustering-Rough Sets-Neural Networks
Yi Fei Yang · Journal of systems management · 2013
Financial crisis warning of listed companies is always an important concern of stakeholders.Due to the data availability,reseachers typically divide companies into two classes as ST and non-ST.Besides,past research pay less attention to the indicator selection,subjective judgement may be the main way for this issue.This paper aims to ovecome the two limitations about finacial situation level and indicator selection.Rough set theory is used to set up a complete and simple indicator system,while hierarchical clustering analysis is used to classify the financial status into five levels,namely health,relative health,medium.Medium,slight warning and serious warning.This changes traditional classification scheme with only ST and non-ST classes.A neural network model is built using the reduced indicator system as the input and the five financial status levels as the output.The model is more accurate in predicting the financial distress status because of more reasonable neural network structure design.