Research on Tourist Flow Analysis and Prediction System Using Big Data

This paper proposes an innovative tourist flow analysis and prediction system, which combines the hybrid algorithm of GNN and LSTM, aiming to solve the limitations of traditional models in processing large-scale, multi-dimensional data. GNN is used to model the spatiotemporal relationship between tourist attractions and surrounding transportation networks, and can effectively capture the complex interactions between scenic spots and between scenic spots and transportation nodes. LSTM is used to capture the time series dependency in passenger flow data, especially long-term dependency features such as seasonal changes and holiday effects. The combination of the two enables the model to simultaneously consider the spatiotemporal relationship and the long-term dependency of historical passenger flow, thereby improving the prediction accuracy. This paper uses large-scale historical passenger flow data, weather data, traffic data and social event data from multiple tourist attractions for training and verification. The experiment results show that the GNN/LSTM hybrid model is superior to the traditional SVM, RF and other single-machine learning models. Compared with the baseline model, the MSE is decreased by 24.7% and 18.5% respectively. Moreover, the system has demonstrated strong real-time forecast ability, and can deal with the high number of concurrent data requests and dynamic adjustment in the scenic area management.