Aiming at the subjective and fixed setting problems of risk assessment index weight in risk access control model of cloud platform, an adaptive weight allocation method of risk assessment index was proposed. Firstly, the the adaptive weight allocation model of risk assessment index was designed through a multivariate linear regression with constraints. Secondly, the programming regression algorithm was proposed and optimized to solve the corresponding weight. Finally, the quantitative formula of risk value with adaptive weight allocation was constructed to calculate the risk value of access request dynamically. The experimental results show that, compared with the Dynamic Risk-based Access Control (DRAC) model and the access control model based on system security risk, the accuracy and sensitivity of risk value of the proposed method are averagely increased by 2.8% and 18.5%, 1.7% and 18.7% with the same order of magnitude training set. Compared with the DRAC model, Dynamic Attribute-based Risk Aware Access Control (DA-RAAC) model and the access control model based on system security risk, the response time of the proposed method is averagely shortened by 9.2%, 34.6% and 96.6% with the same number of access requests. The proposed method has higher accuracy and sensitivity in the risk value of large concurrent users, and its response time is shorter, which is more suitable for cloud environment.