An Effective LSTM Model for Predicting Tourist Flow at Scenic Spot Based on Seasonal-Trend Decomposition and Discrete Cosine Transform

In order to solve the problem of uncertainty and randomness in data mining and feature extraction of tourist flow prediction model, a new hybrid model 'STL-DCT-LSTM' is proposed in this paper. In this model, the seasonal trend decomposition algorithm is first used for data preprocessing, and the tourist flow data of scenic spots is decomposed into three feature channels to extract the feature-related information. Then Fourier transform algorithm is used to convert the time domain information of tourist flow into frequency domain information, and discrete cosine algorithm is used to simulate the frequency correlation between feature channels, so as to alleviate the Gibbs phenomenon of high-frequency noise introduced in the Fourier transform process of Fourier algorithm, improve the ability of the model to extract frequency features and automatically optimize the weight and threshold of each feature. Finally, the long and short term memory neural network is used to forecast the tourist flow of the scenic spot, and the components of the forecast result are added by their respective weights to get the final predicted value. In addition, the STL-DCT-LSTM model is compared with other predictive modeling methods such as linear regression, regression tree and support vector machine, and the results show that the STL-DCT-LSTM model is the best predictive model.