Research on multiview clustering algorithm based on deep self-encoder and co-training

In the era of big data, clustering analysis is widely used in various fields as an important data exploration tool. However, traditional clustering algorithms usually rely on data from a single view and fail to effectively integrate information from different views, thus limiting the clustering effect. To solve this problem, this paper proposes a multi-view clustering algorithm (PDEMVC) based on deep self-encoder and co-training. The PDEMVC algorithm first learns the features of each view through the deep self-encoder to obtain the embedded feature representation. Subsequently, the clustering result of one view is used as a reference to guide the clustering process of other views using the co-training mechanism. This approach aims to improve the overall performance of multi-view clustering. Experimental results show that PDEMVC achieves significant clustering results on multiple datasets (e.g., MNIST-USPS, Fashion, and BDGP), with significant improvements in accuracy and Normalised Mutual Information (NMI).PDEMVC effectively solves the multi-view clustering problem, enhances the clustering accuracy and robustness, and, compared to the existing methods, exhibits higher accuracy and NMI values. This study provides a new idea and an effective method for clustering analysis of multi-view data.