Prediction of tourist flow based on multi-source traffic data in scenic spot

At present, China's tourism industry has entered an era of rapid development. It is of great significance to predict the precise and real-time tourist flow of the popular tourist areas with dense crowds, providing decision support for rational evacuation of tourist flow, activation of emergency plans, and prevention of security accidents. This article presents a long short-term memory (LSTM) neural network model based on the global attention mechanism. The model first uses two LSTM layers to calculate attention weights using multi-source data at different time steps to improve the model's learning ability. Then the model predicts the tourist flow of the scenic spots in the next time period based on the weights and outputs previously obtained. At the final stage, the pre-trained parameters of the network are used to initialize the model. To verify the validity of the model, we compared it with the LSTM model, back propagation neural network model, and autoregressive integrated moving average model based on the data of Beijing's South Luogu Lane scenic spot. It turned out that the results of our solution were more in line with the true value, which proves that it is feasible for real-time prediction of tourist flow in popular scenic spots.