Insights from Tweets: Analysing Destination Topics and Sentiments, and Predicting Tourist Arrivals

Abstract

Social media has gained great popularity among consumers and social media data offers enormous potential for marketing researchers to generate consumer insights. This thesis attempts to examine how tweets can be analysed and used to predict tourist arrivals. The research objectives are: 1) to identify the most important topics from tweets; 2) to quantify the sentiments of the topics extracted; 3) to examine how the topics can predict tourist arrivals.
Two large and popular destinations were used for empirical analysis. Study 1 focuses on Sydney, Australia and Study 2 focuses on London, the UK. First, tweets mentioning the destination were extracted and tourist arrivals data were retrieved from official sources. A topic modelling algorithm, BERTopic, was then used to extract important topics from the tweets data. The sentiment scores of each topic then were calculated. Finally, H2O AutoML was used to the forecast model of tourist arrivals using the sentiment scores. The best performing model for the destination was selected. The relationships between different topics and tourist arrivals are then identified.
The results indicate that people pay attention to different topics for the two destinations. For Sydney, the main topics are events and activities and food, while for London, the main topics are events and activities, and tourism facilities. The influential Twitter topics for predicting tourist arrivals are also different. For Sydney, these are events and activities, travel costs, food, and symbolic factors, while for London, food, symbolic factors, internal factors, and geographic factors are the important factors.
The thesis extends the marketing research literature by identifying influential topics for tourist arrivals from social media, and how the sentiments on different topics discussed on social media influence tourist arrivals. It also makes methodological contributions and has important implications for practice.