<p>Deep multi-view clustering (DMVC), with its ability to effectively mine cross-view feature representations, has garnered significant attention in recent years. At present, many deep learning-based multi-view clustering methods have been proposed and have achieved good clustering performance. However, most of the existing methods fail to fully utilize the semantic information of samples and are susceptible to the influence of view-private information, making it difficult to learn discriminative representations with clear clustering structures, thereby limiting the improvement of clustering performance. To address the aforementioned issues, this paper proposes a novel Deep Multi-View Clustering via Dual Contrastive Consistency Fusion (DCCF). Firstly, to obtain fusion features with a clear clustering structure, a clustering-guided fusion module was designed that leverages consistency information in the semantic space as a self-supervised signal, enabling samples with similar semantics to be close to each other in the feature space. Subsequently, through dual contrastive learning, the consensus representation and view-specific representations are aligned. Eventually, the feature fusion process is fine-tuned using the optimized global semantic information of the sample to mitigate the impact of view imbalance. Extensive experiments on nine public datasets demonstrate that DCCF outperforms state-of-the-art methods in clustering performance. The code is available at <a href="https://github.com/zhaole1204/DCCF.">https://github.com/zhaole1204/DCCF.</a></p>