In order to effectively control spam SMS and continuously improve the accuracy of the management

this paper constructs a scientific hierarchical spam message management system based on XGBoost algorithm. By modeling and calculating the characteristics of users’ network access

consumption and communication behavior

the risk coefficient of users sending spam messages is identified

and six levels of users are defined. After the hierarchical list is generated

the corresponding interception strategies are set respectively based on the hierarchical policy management function of the spam message interception system. While strengthening the identification and interception of medium and high-risk users

it effectively reduces the risk of false interception of high-value users due to bad information governance. Since the application of the governance system to some operator

the average number of spam SMS complaints has decreased by 34.82%. The complaint rate of false shutdown is stable within 0.1%. The number of shutdown complaints from tier 1 to tier 3 high-value customers decreased by 27.43%.