Tourism demand forecasting based on multi-source fine-grained sentiment mining
In the digital economy era,the content generated by consumers based on internet social media platforms,such as searches and reviews,has expanded the data sources for tourism demand prediction.Existing research focuses on mining and analyzing consumers'overall emo-tional tendencies towards tourist destinations,but there is little consideration of the impact of their differentiated evaluations on fine-grained levels such as dining,accommodation,transporta-tion,and services on tourism demand prediction.Taking Jiuzhaigou scenic area's visitor flow prediction as an example,this paper uses machine learning-based fine-grained sentiment analysis methods to analyze multi-source data from Ctrip,Qunar,Dianping,and Meituan,constructing a fine-grained sentiment index covering eight dimensions to predict the scenic area's visitor flow.This is compared with models containing search engine indices and overall sentiment indices.The results show that,in three different types of prediction models based on time series,machine learning,and deep learning,the models containing the fine-grained sentiment index can signifi-cantly improve the accuracy of tourism demand prediction.In the out-of-sample forecasting,the model incorporating a fine-grained sentiment index demonstrates an average prediction accuracy improvement of 17.78％and 6.53％compared to models that include search engine indices and a general sentiment index.This study provides innovative research methods for multi-dimensional data-driven tourism demand prediction in the era of big data.
fine-grained sentiment miningmulti-source big datatourism demand forecastingaspect-based sentiment analysis
NETL
NSTL
万方数据
