Multi-Regularized Nonnegative Tensor Factorization for Context-Aware Recommendation

As a powerful high-dimensional data analysis tool, tensor factorization has important applications in context-aware recommendation systems. However, traditional tensor factorization methods often face problems of model overfitting and low recommended accuracy when dealing with high-dimensional sparse data. To overcome these challenges, this paper proposes a Multi-regularized Nonnegative Tensor Factorization (MNTF) model that combines $L_{1}$ and $L_{2}$ regularization. By introducing $L_{1}$ regularization, the model can achieve sparsity, thereby effectively selecting important features and reducing complexity. At the same time, $L_{2}$ regularization further improves the generalization ability of the model by smoothing parameter values. In addition, the nonnegative constraint ensures that the decomposition results have better interpretability, making them more consistent with the data characteristics in actual application scenarios. Experimental results show that compared with other current methods, this method significantly improves the recommendation performance on multiple real datasets.