Research On Forecasting Method Of Tourist Number Based On Echo State Network

With the continuous development of the tourism industry,accurate prediction of tourism demand is an important prerequisite and guarantee for the rapid and steady development of the tourism industry.Due to the limitation of tourism cycle and capacity,the supply and demand of tourism resources will be out of balance.Optimizing the allocation of tourism resources requires the establishment of a reasonable prediction model to provide data support for the distribution of tourism resources.Therefore,accurate prediction of tourism demand is of great significance to the development of tourism.This article focuses on the research on the number of tourists in the tourism demand.The specific work can be divided into the following parts:This thesis studies the problem of the number of tourists arriving.The specific work can be divided into the following parts:Aiming at the problem that the reserve pool in the echo state network is randomly generated,there is a strong coupling between neurons,and the prediction accuracy is affected,an echo state network model based on small-world characteristics is designed.In addition,the sigmoid function of the neuron in the echo state network is prone to zigzag phenomenon,and it is also prone to produce a large amount of redundant information.The tanh function and wavelet function are selected as the neuron excitation function in the echo state network at the same time,which increases the transformation of the neuron.The form improves the state space of the echo state network.Aiming at the timing and non-linear characteristics of the tourism prediction problem,the complex internal structure of the traditional echo state network reserve pool leads to higher output dimensions,collinearity between the data,and even over-fitting problems,so a fusion auto-encoding is designed The predictive model of the echo state network and the echo state network uses the autoencoder to reduce the dimensionality of the output of the reserve pool,thereby improving the prediction accuracy.The simulation verifies that the prediction performance of the model is better than the traditional echo state network.The internal structure of the echo state network for small-world characteristics will become more complex,and the output dimension will be higher.The echo state network based on self-encoding can reduce the output dimension and improve the prediction accuracy.Based on this,a self-encoding-based The network prediction model of the small-world echo state of the device.The paper applies this model to the problem of tourist arrivals prediction,and uses the number of tourists in Singapore from 2000 to 2018 as a data set for simulation experiments to verify that the modelhas higher prediction accuracy than traditional models such as echo state networks.At the same time,using this model with autoregressive models,support vector regression models,and neural network models to predict and compare the number of tourists from ten countries to Turkey,the results prove the superiority of the model.The improved model based on the echo state network proposed in the thesis helps the tourism industry to formulate reasonable plans and measures.While accurately predicting the number of tourists,it can optimize the allocation of tourism resources and provide a more reliable development for the tourism industry.Forecast model.