مقاله انگلیسی یش بینی ورود گردشگران با یادگیری ماشین (2019 الزویر)

Previous studies have shown that online data, such as search engine queries, is a new source of data that can be used to forecast tourism demand. In this study, we propose a forecasting framework that uses machine learning and internet search indexes to forecast tourist arrivals for popular destinations in China and compared its forecasting performance to the search results generated by Google and Baidu, respectively. This study verifies the Granger causality and co-integration relationship between internet search index and tourist arrivals of Beijing. Our experimental results suggest that compared with benchmark models, the proposed kernel extreme learning machine (KELM) models, which integrate tourist volume series with Baidu Index and Google Index, can improve the forecasting performance significantly in terms of both forecasting accuracy and robustness analysis.

Introduction

All over the world, the tourism industry contributes significantly to economic growth (Gunter & Onder, 2015; Song, Li, Witt, & Athanasopoulos, 2011). According to the China National Tourism Administration, in 2016 the tourism income of China reached 4.69 trillion RMB, increasing by 13.6% compared to the previous year, and accounted for 6.3% of China’s GDP. Thus, forecasting tourist volume is becoming increasingly important for predicting future economic development. Tourism demand forecasting may provide basic information for subsequent planning and policy making (Chu, 2008; Witt & Song, 2002). Methods used in tourism modeling and forecasting fall into four groups: time series models, econometrics models, artificial intelligence techniques and qualitative methods (Goh & Law, 2011; Song & Li, 2008). In addition to simple tourist data announced by the State Statistics Bureau, Internet search queries, which reflect the behavior and intentions of tourists, have increasingly been used in tourism forecasting models (Croce, 2017; Goodwin, 2008). However, the search index has created big opportunities in the modeling process of tourism forecasting (Li, Pan, Raw & Huang, 2017). Internet search data has been applied to many aspects, such as hotel registrations (Pan & Yang, 2017; Rivera, 2016), tourist numbers (Bangwayo-Skeete & Skeete, 2015; Yang, Pan, Evans, & Lv, 2015), economic indicators (Choi & Varian, 2012), unemployment rates (Askitas & Zimmermann, 2009), private consumption (Vosen & Schmidt, 2011), and stock returns (Zhu & Bao, 2014). When introducing the Baidu Index or Google Index into forecasting models, keywords and the composition of indexes must be selected carefully. Keywords can be selected according to the correlation coefficient, the tendency chart or the crowd-squared method (Brynjolfsson, Geva, & Reichman, 2016). Additionally, the composition of indexes can be achieved by the HE-TDC method (Peng, Liu, Wang, & Gu, 2017) or the principal component analysis (PCA). Obviously, efforts should be made to avoid problems related to multi-collinearity and over-fitting to the greatest extent possible.