Abstract: The regional cooperation between China and the countries involved in the Belt and Road Initiative (BRI) is deepening, and China is increasing its investment in these countries. However, China is different from these countries in terms of political system, social development and cultural heritage, and the social risk exposure of these countries is also different. With the help of the machine learning method of big data analysis techniques, the paper constructs a BP neural network based on social risk of the countries involved in the BRI, and measures the social risk level of these countries with relative accuracy. An assessment of the risk data of the 50 countries involved in the BRI from 2010 to 2017 shows that most of these countries are at high levels of social risk, accounting for 46% of all the countries evaluated, and only some countries in Eastern Europe and a few countries in East Asia, Southeast Asia and Oceania are at low levels of social risk. Even for countries in the same geographical region, their levels of social risk also show significant differences.