<p>Multi-view clustering is a significant technique within the realm of machine learning, designed to uncover the underlying structure of data by conducting an integrated analysis of multi-source data. However, current multi-view clustering methods exhibit three principal shortcomings. Firstly, existing approaches seldom apply filters to smooth data representations. Secondly, traditional multi-view clustering methods struggle with the effective integration of information from diverse views, leading to inadequate utilization of information and challenges in addressing incompleteness and inconsistency between views. Thirdly, most methods do not fully exploit high-order structural information among data. To address these limitations, this paper proposes a novel method termed high-order tensor based multi-view clustering via enhanced adaptive graph propagation. Specifically, the method utilizes Gaussian filters to smooth data representations, thereby mitigating noise and enhancing the quality of the data. An enhanced adaptive graph propagation mechanism is then implemented to intensify interaction and information exchange among nodes, and facilitates the collaboration and fusion of information across different views. Simultaneously, high-order tensors are utilized to integrate multi-view data, accurately reveal the underlying structure of the data, and fully exploit the similarity information of high-order structures among the data. Experimental results demonstrate the superiority of our method over other clustering methods across seven real-world datasets.</p>