The cash flow early warning indicators system is established according to the current research dynamic at home and a- broad and the financial crisis early warning perspective. 79 ST and 79 non-ST companies of Shanghai and Shenzhen stock markets from 2007 to 2012 are selected as the training example. The 3-year statistics of the training example before they are ST are used to select in- dicators. The prediction result of BP neural network model is compared to the truth according to the selected indicators and BP neural network model. The researc...