This paper uses Logistic regression to establish a financial crisis warning model based on accounting anomalies indicators, financial indicators and non-financial indicators, it makes 224 companies listed on the Shanghai and Shenzhen Stock Exchanges as a research object from 2017 to 2019, including 112 companies in a special trade and that much in normal trade, and it tests the effectiveness of the warning model. The results show that the accounting anomalies indicators can effectively improve the predictive ability of the financial crisis warning model, and it has important reference value for investors to effectively identify financial crisis and government departments to strengthen the supervision of listed companies.