Abstract: To address the problem of poor fusion effectiveness at the feature level due to the dimensional disparity among multi-omics data, resulting in low model classification accuracy, this paper proposes a multi-omics data fusion classification method based on hypergraphs. For datasets with lower-dimensional features, a matrix concatenation approach is employed to fuse different omics data, while for datasets with higher-dimensional features, a fusion distance matrix is utilized to integrate multiple omics data. The fused data contain correlation information between omics, facilitating enhanced model learning performance. A Weighted Hypergraph Neural Network model (WHGNN) with weight representation is designed based on the hypergraph neural network. During the convolution process, vertex weight matrices and hyperedge weight matrices are introduced to represent the importance of nodes and hyperedges, respectively. The hypergraph structure enables the transmission of high-order neighborhood information of nodes and the learning of hyperedge structure features, while the weight matrices help mitigate overfitting and enhance the learning of features of important nodes and hyperedges, thereby improving the model's classification performance. Experimental results demonstrate classification accuracies of 81.9%, 87%, and 89% on publicly available datasets of Alzheimer's disease, breast cancer cells, and glioblastoma cells, respectively, outperforming classical classification algorithms such as KNN, RF, and SVM.