As the communication and origin of online public opinion, social platforms play an extremely important role in guiding public opinion. Therefore, strengthening the supervision of social media will become one of the necessary means to prevent losing control of public opinion. For better guide public opinion, fundamentally explore the impact of the strength of social media supervision on the network public opinion. This paper constructs an evolutionary game model between network platform and Internet users, and analyzes the game relationship between Internet users ( publishing false information and publishing correct information) under the supervision of platform punishment mechanism and no platform punishment mechanism. The study finds that when the network platform takes strict supervisory measures, most netizens will choose to publish correct information. When the network platform supervisory measures are weak, most netizens will choose to publish false information in order to obtain higher benefits. This paper suggests that when network public opinion break out on a platform, the platform should take strict regulatory measures to control and guide them. Therefore, it is suggested that the platform should establish or optimize the user punishment mechanism, monitor the background data information status in real time, control the transmission channels and directions of public opinion in the early stage, and enhance the controllability of public opinion events. The research conclusions and suggestions support the development trend of contemporary social media supervision, and provide a theoretical basis for social platforms to cope with the complex and volatile crisis of public opinion.