This study introduces a novel hybrid framework, termed Mamba-iTransformer, designed for tourism demand forecasting. The framework synergistically integrates the linear time complexity and long-range dependency modeling capabilities of Mamba with the multivariate correlation capturing and non-linear representation learning strengths of iTransformer. By leveraging this integration, Mamba-iTransformer adeptly captures both long- and short-term dependencies, thereby enhancing the accuracy of time series forecasting while maintaining computational efficiency in processing extended sequence data. Empirical evaluations reveal that Mamba-iTransformer achieves superior predictive performance across multiple real-world datasets, significantly outperforming existing state-of-the-art methodologies. This research contributes a highly efficient and precise approach to tourism demand forecasting, demonstrating robust generalization capabilities and adaptability.