<p>Resource allocation is an essential aspect of many fields, such as loan management, cloud computing, and advertising, where making optimal decisions by minimizing total costs under resource consumption constraints is critical. However, traditional resource allocation models bring about several challenges, such as high computational complexity, inadequate performance on test datasets, and insufficient robustness in yield generation. To overcome these challenges, we propose a novel Wasserstein distance based distributionally robust resource allocation model by employing a sample grouping strategy. This approach improves performance on test datasets and reduces the number of variables. The proposed model is reformulated as a linear programming with regularization terms. Numerical results demonstrate that the proposed model achieves superior generalization performance and robustness compared to the sample-average approximation model and the distributionally robust resource allocation model without a sample grouping strategy. Moreover, the computing time is significantly reduced compared to those methods without the grouping strategy.</p>