Research On Tourism Demand Forecasting Based On Machine Learning

Most tourism products are perishable and cannot be stored for a long time.Moreover,the distribution of tourism resources is not balanced with the actual number of tourists,which leads to the waste of a large number of tourism resources.Therefore,it is very important to predict the accurate tourism demand.This article studies the forecast of tourism demand,and then puts forward the tourism forecast model to improve the prediction accuracy.The main research work is as follows:1.In order to solve the problem that traditional machine learning models are difficult to mine the complex relationships in multivariate tourism data,an autoencoder based on LSTM prediction model,namely AE-LSTM model,is proposed.This model can be divided into the pre-training stage and fine-tuning stage of the autoencoder based on LSTM.Based on the AE-LSTM model,The Stacked autoencoders based on LSTM are deeply Stacked to ensure that the hidden layer parameters trained by each autoencoder are locally optimal.Thus,the Stacked autoencoder based on LSTM prediction model,namely SAE-LSTM model,is proposed.2.In view of the shortcoming that LSTM will still lose part of the tourism data when the input sequence is too long,the Attention Mechanism is introduced on the basis of SAE-LSTM model to build the Attentional-SAE-LSTM prediction model,which has the advantage of being able to extract important tourism information.3.The model proposed in this article was applied to the Macao dataset,and LSTM and stacked LSTM were added as the benchmark model.The experiment proved that the prediction effect of the AE-LSTM and SAE-LSTM model was better than the benchmark model,and the prediction effect of the Attention-SAE-LSTM model was better than the AE-LSTM and SAE-LSTM model.In addition,attention score is used to analyze the influencing factors of tourism.This article proves the predictive ability of the proposed Attention-SAE-LSTM model in tourism volume prediction through experiments,which helps the relevant tourism departments to understand the distribution of passenger flow in advance and make scientific decisions.