A prediction study on tourist amount based on web search data — A case from Hainan

The web search data, which records hundreds of millions of searchers' concerns and interests, reflects the trends of their behavior and provides essential data basis for the prediction of tourist amount. In this paper, firstly, a systematic theoretical framework has been built to reveal the correlation between web search and tourists' travel. Secondly, at the theoretical framework' basis, an empirical study on Hainan verified the co-integration relationship between search index and tourist amount. Finally, an prediction model has been established to predict consecutive 4 months' Hainan tourist amount. The results show that compared with the traditional auto-regression AR model, adding search index model's Mean Absolute Percent Error( MAPE) decrease from 6.54% to 1.34%, goodness of fit reaches 0.975, and realizes “predict the present” making up China National Tourism Administration's(CNTA) data release delay for about 1 month, confirming the prediction ability of search index for tourist amount. This paper's conclusions can be provided as references for CNTA monitoring the change of tourist amount and tourism service offering adequate ancillary services. The new prediction method considering search index can also be applied to other web-based soc-economical activity.