The financial distress of listed companies not only threatens the interests of the enterprise and internal staff, but also makes investors face significant financial loss. It is important to establish an effective early warning system for prediction of financial distress. Financial news contains a lot of unstructured text data about the financial status of the business. Therefore, this paper takes into account the unstructured text data to the early warning system for sentimental analysis. In the section of financial indicators, this study uses the seven major categories of financial ratios in the ZETA model of Altman (2000). We use logistic regression and random forest to establish our model. However, the weakness of ZETA model is that the prediction accuracy will be greatly dropped over two years. This study introduces a hidden Markov model to improve the long-term prediction accuracy of the model. This paper provides a hybrid method which integrates text mining and hidden Markov model for prediction of financial distress for listed companies in Taiwan.