With the increasing personalization and diversification of user needs, traditional cloud storage services have exposed prominent problems such as declining user awareness, convergence of service products, and low marketing efficiency. Therefore, this article focuses on the analysis of user behavior characteristics and dynamic optimization of storage resources, and designs a user clustering method that integrates an improved behavior indicator system and fuzzy clustering optimization algorithm. In terms of customer value characterization, an RFAU indicator framework is constructed to incorporate user participation frequency and consumption activity into the evaluation dimensions, and a multi-level initialization strategy is introduced to improve the fuzzy C-means clustering method (HFCM-DPC) to enhance the discriminability and adaptability of clustering. The test results show that the model can more effectively explore the potential value of users, and is significantly better than conventional methods in clustering accuracy and convergence performance. To adapt to the storage requirements of user behavior evolving over time, a replica management mechanism oriented towards changes in access behavior is further proposed. Experimental verification shows that this scheme not only improves the response speed of storage access, but also exhibits good dynamic balancing ability in system load distribution. The research results of this article are expected to provide theoretical support and feasible strategies for differentiated service operation and resource elastic scheduling of cloud platforms.