Tensor joint framework enhanced by game-theoretic constrained optimization for skin cancer forecasting

Skin cancer is a public health concern due to its high incidence and detection challenges. While tensor decomposition is widely utilized to predict miRNA-disease associations, existing models are not optimized for skin cancer, thereby limiting their comprehensiveness and interpretability. To address these limitations, we propose a many-objective tensor joint framework aimed at enhancing prediction accuracy while maintaining comprehensiveness and interpretability. The proposed framework classifies tensors according to their similarity to skin cancer and employs a many-objective optimization algorithm to optimize weight allocation. Similarity constraints for skin cancer are applied to prioritize relevant information, effectively minimizing noise from unrelated diseases. Furthermore, we introduce a multi-stage constrained many-objective optimization algorithm based on game theory. This algorithm leverages game theory to dynamically adjust population diversity, convergence, and constraint violations throughout the evolutionary process, thereby improving the overall framework's performance. Experimental results demonstrate that the proposed algorithm outperforms existing state-of-the-art constrained many-objective optimization algorithms on benchmarks such as MW and DTLZ. Compared to other tensor decomposition methods, the proposed framework achieves a comprehensive improvement ranging from 8.49% to 11.53% across four objectives.