Multi-view clustering has a broad range of real-world applications due to its ability to process data from multiple sources. However, these data often have noise and missing cases that most multi-view clustering methods don’t care about. These methods may be challenging to employ directly due to the absence of instances, and the presence of noise will result in unreliable clustering results. In order to resolve these obstacles, this article proposes a novel incomplete Multi-view Anchor Graph Fuzzy System to solve the incomplete multi-view data. The initial step is to construct a multi-view dual anchor graph in order to acquire information from the original view and hidden space. Then, a cooperative fuzzy clustering is suggested to further investigate and make use of both the common and specific information based on the embeddings taken from dual anchor graphs. Moreover, the weights of different views are learned using a fuzzy membership structure preservation approach combined with negative Shannon entropy to attain optimal clustering performance. Finally, the inference rule and clustering outcomes are combined to impute missing values in the multi-view data. Experimental results on real-life educational datasets illustrate the effectiveness of the proposed approach.