基于ARIMA和LSTM神经网络对中国入境游客规模预测的比较研究

A Comparative Study on Forecasting the Size of Chinese Inbound Tourists Based on ARIMA and LSTM Neural Network

摘要: 随着我国综合国力的全面提升，我国旅游业也进入高速发展阶段，入境旅游人数日益增多，准确预测我国入境游客规模具有重要意义。本文分别使用LSTM (Long Short-Term Memory)模型和ARIMA (Autoregressive Integrated Moving Average Model)模型对我国入境游客人数进行预测对比，并以2014年1月至2016年12月的我国入境游客人次为例，进行实证研究。结果都表明LSTM神经网络比ARIMA更适合我国入境游客规模预测，LSTM模型预测精度比ARIMA高22.981%。基于LSTM模型预测入境游客人数，对相关部门优化旅游资源配置，具有一定的指导意义。

Abstract: With the overall improvement of China’s comprehensive national strength, China's tourism in-dustry has entered a stage of rapid development. The number of inbound tourists is increasing. It is of great significance to accurately predict the scale of inbound tourists in China. This paper uses LSTM (Long Short-Term Memory) model and ARIMA (Autoregressive Integrated Moving Average Model) model to predict and compare the number of inbound tourists in China, and takes the number of inbound tourists from January 2014 to December 2016 as an example to conduct empirical research. The results show that LSTM neural network is more suitable than ARIMA for predicting the scale of inbound tourists in China, and the prediction accuracy of LSTM model is 22.981% higher than ARIMA. Predicting the number of inbound tourists based on LSTM model has certain guiding significance for relevant departments to optimize the allocation of tourism resources.