Abstract: With great economic downward pressure, many companies are facing performance thunder frequently. How to identify and guard against ‘traps’, reshape investors' confidence and prevent financial crisis has become a practical and urgent task. With the theoretical framework of TEI@I methodology, this paper integrates text mining and deep learning to construct a prediction model of enterprise financial crisis. From a new perspective, a dynamic modeling method for financial distress prediction which is based on convolutional neural networks and long-term and short-term memory networks is proposed. The empirical research is carried out with listed Chinese companies which in the information service industry as samples. With the effective considering of financial factors and non-financial factors that affect financial crisis of enterprises, the forecasting model has much better forecasting effect than others. This method can be applied and promoted easily, and has positive value for government and investors if they need to grasp the business dynamics, reduce investment risks and prevent financial crisis.

Key words: financial risk warning, TEI@I methodology, text mining, convolutional neural network, short-term memory network