Breast cancer is a heterogeneous disease characterized by molecular complexity. This research utilized three genetic expression profiles—gene expression, deoxyribonucleic acid (DNA) methylation, and micro ribonucleic acid (miRNA) expression—to deepen the understanding of breast cancer biology and contribute to the development of a reliable survival rate prediction model. During the preprocessing phase, principal component analysis (PCA) was applied to reduce the dimensionality of each dataset before computing consensus features across the three omics datasets. By integrating these datasets with the consensus features, the model's ability to uncover deep connections within the data was significantly improved. The proposed multimodal deep learning multigenetic features (MDL-MG) architecture incorporates a custom attention mechanism (CAM), bidirectional long short-term memory (BLSTM), and convolutional neural networks (CNNs). Additionally, the model was optimized to handle contrastive loss by extracting distinguishing features using a Siamese network (SN) architecture with a Euclidean distance metric. To assess the effectiveness of this approach, various evaluation metrics were applied to the cancer genome atlas (TCGA-BREAST) dataset. The model achieved 100% accuracy and demonstrated improvements in recall (16.2%), area under the curve (AUC) (29.3%), and precision (10.4%) while reducing complexity. These results highlight the model's efficacy in accurately predicting cancer survival rates.