Title: Forecasting tourist arrivals with the help of web sentiment: A mixed-frequency modeling approach for big data
Authors: Irem Onder - MODUL University Vienna (Austria)
Ulrich Gunter - MODUL University Vienna (Austria) [presenting]
Arno Scharl - MODUL University Vienna (Austria)
Abstract: Online news media coverage of a destination can affect destination image and, in turn, can influence the number of tourist arrivals. It is a form of Big Data, which can be crawled and collected in various ways. Moreover, destination image can change during and after an online browsing session when reading news related to a destination. Sentiment analysis extracts web sentiment by rating a segment of text as either positive (favorable) or negative (unfavorable), which shows the perception of the news author about a topic in discussion. The goal is to investigate whether web sentiment data, which are based on online news media coverage of four European cities (Berlin, Brussels, Paris, and Vienna), possess informative content able to predict actual tourist arrivals. To achieve this goal, sentiment analysis of online news media coverage was conducted using automated semantic routines. Due to different data frequencies of tourist arrivals (monthly) and web sentiment indicators (daily), the mixed-data sampling (MIDAS) modeling approach was applied. Results show that MIDAS models employing various types of web sentiment indicators are able to outperform time-series and naive benchmarks in terms of typical accuracy measures.