多源异构数据融合的旅游规模预测研究

Forecasting tourism scale using multi-source heterogeneous data fusion

旅游规模预测研究亟待应用旅游大数据及其适应性方法,以提升预测的准确度。文章基于多来源搜索引擎指数和UGC数据,以ARIMA模型为基准,综合运用灰色关联度分析、TF-IDF自然语言处理技术、机器学习领域中适用范围最广的RF与SVR模型,对重庆市月度旅游规模展开预测。研究发现:融合多来源搜索引擎指数与包含旅游者丰富心理信息的UGC数据进行的旅游规模预测,在三种模型中的预测精度均呈提升态势,说明多源旅游大数据对目的地潜在客流具有前兆作用。在相同数据集约束下,ARIMAX、RF、SVR三种模型的预测精度递增,表明机器学习算法对于旅游大数据具有更强的分析与处理能力。文中提供了集成多源异构旅游大数据与高效的机器学习算法以回应城市旅游规模预测问题的新思路。

It is urgently needed to apply tourism big data and its adaptive methods to forecasting tourism scale for improving the accuracy of forecasting.Based on multi-source search engine index and user generated content data(UGC),the ARIMA model is taken as a benchmark,gray correlation analysis,TF-IDF technology,and both of support vector regression and random forest method which are the most widely applicable in the field of machine learning,are comprehensively used to forecast the monthly tourism scale in Chongqing.The following two main results are obtained.Firstly,the forecasting accuracy using multi-source search engine index data and UGC data containing tourists’rich psychological information shows a trend of improvement in each of three models(i.e.,ARIMAX,RF,SVR models).The multi-source tourism big data has a precursory effect on the potential scale of the destination.Secondly,under the same dataset constraints,the forecasting accuracy of ARIMAX,RF,and SVR models are gradually improved,indicating that machine learning algorithms have stronger analyzing and processing capabilities for tourism big data.This paper attempts to forecast the scale of urban tourism by a new research path,which integrates multi-source heterogeneous tourism big data and efficient machine learning algorithms.