Multi-model Fusion Network for Tourism Prediction

Tourism prediction plays a crucial role in the vibrant development of the tourism industry. As the complexity of tourism data continues to increase, the potential features contained within tourism data also become abundant. Traditional prediction methods struggle to exploit the latent features within tourism data, as well as the temporal and spatial characteristics of the data simultaneously. To address this issue, a spatiotemporal network (ST-Net) prediction model that combines multiple models is proposed. Initially, the dimension of the data is enhanced using empirical mode decomposition, increasing the spatiotemporal features of 1D tourism data. Then, the spatiotemporal graph convolutional network and gated recurrent unit models are employed to accurately extract tourism features, further incorporating the temporal and spatial feature information of the tourism data. Finally, the transformer and long short-term memory models are used collaboratively for regressive prediction in tourism. Extensive experimental results demonstrate that the proposed ST-Net prediction model is more accurate, efficient, and stable than similar approaches.