The study applies machine learning algorithms to the national economy, and selects one of the Extreme Learning Machine algorithms to be embedded in the field of financial risk prevention and control. The financial risk monitoring model and the financial risk warning model based on Extreme Learning Machine (ELM) are constructed to prevent and control financial risks. The ELM model is compared with other early warning models in terms of prediction performance to get a comprehensive evaluation of the ELM model. And then, the SHAP explanatory model is utilized to measure the degree of importance and impact of each financial risk warning feature. The ELM model is used to monitor China’s financial risks from 2008-2021, analyze China’s financial stress, and predict the possibility of China’s outbreak of systemic financial risks in 2022-2023.The overall accuracy of the ELM algorithmic model is 0.990, which exceeds that of other early warning models. Among the top ten characteristic indicator variables in terms of importance, the closing price, the maximum price and the interbank 7-day pledged repo weighted interest rate are the indicators that pull the probability of risk warning, and the SZSE Composite Index, the S&P 500 Index, the foreign exchange reserves, the year-on-year growth rate of M2, and the Nikkei 225 Index are the indicators that reduce the probability of risk.In 2008-2013, the systemic financial risk was in the basic safety zone.In 2014- 2015 is in the alert zone. 2016-2017 is in the basic safety zone. 2018 is in the safety zone. 2019 is in the near-alert zone. 2020 enters the danger zone. 2021 is in the basic safety zone. 2022-2023 has a low probability of the outbreak of systemic financial risk.
