China's Xinjiang region has gradually explored a successful experience of “de-extremization” to remove the ideological roots of terrorism and extremists by “combining attack and prevention， focusing on prevention”. In accordance with the relevant laws and regulations， through the establishment of vocational skills education and training center in Xinjiang， the source control of de radicalization for the groups participating in terrorist activities and extremist activities is realized. Due to the different degree of influence of extreme thoughts on trainees， their behavioral motivations for training are also quite different. The degree of cooperation of trainees is affected by the efficiency and effect of vocational skills training， the degree of fanaticism of extreme thoughts and the coercion of overseas religious extremist forces. However， this successful anti-terrorism experience has been maliciously discredited by the United States and other western countries. For the first time， the successful experience of “de-extremization” activities such as establishing vocational skills education and training centers in accordance with the law in Xinjiang is systematically demonstrated.According to the decision-making motivation of three different types of students， such as mild extreme behavior， severe extreme behavior and being coerced by overseas extremism， the expected return function is expanded into irrational factors and external random interference factors respectively. The stochastic evolutionary game model of limited group is established， and the dominant condition of “active cooperation” strategy becoming evolutionary stable strategy in student group is discussed. In this paper， Hetian vocational skills training center as a case， the relevant data and content are from the “vocational skills education and training in Xinjiang” white paper and other public information channels.The research shows that the strategy evolution of trainees mainly depends on the effect and efficiency of training， the development benefits of eliminating extremes and the scale of training center. For the students with mild extreme behavior， the positive cooperation strategy needs to meet certain income conditions to become the evolutionary stable strategy. For the students with severe extreme behavior， the negative treatment strategy is always an evolutionary stable strategy. When the external random interference factors occupy the dominant position， these interference factors will seriously hinder the de extreme teaching and training work. The above research results show that the combination of economic motivation means to improve the training effect and efficiency with non economic motivation means such as criminal investigation， intelligence interception and training scale is conducive to the development of de-extremization.The following problems have not been considered in this paper. First of all， there is no in-depth discussion on the infiltration process of overseas extremism， that is， the types of external interference factors and the ways to deal with them. The incentive of students' behavior is often the entanglement of economic interests， ideology and external random factors. Secondly， in the aspect of data analysis， it is helpful for the international community to formulate anti-terrorism de-extremization measures according to local conditions by analyzing the real behavior characteristics of students based on big data. Third， in the practice of international anti-terrorism de radicalization， countries are constantly exploring the source governance path in line with their own national conditions. In the future research， a comparative analysis can be made from a horizontal perspective， and strong theoretical proof and guidance can be provided for China's anti-terrorism “de-extremization” work in the next step.