Evaluating the Effect of Weather on Tourist Revisit Intention using Natural Language Processing and Classification Techniques

Tourists’ revisit has significant monetary benefits to destinations because the cost of retaining existing visitors is less than attracting new visitors. Re-visit intention is often based on tourists experience and satisfaction at a destination. An important aspect that influences the relationship between satisfaction and intention to revisit is the weather conditions at a destination given the increased frequency of heatwaves that strike summer holiday destinations over the summer months. This work applies natural language processing and classification techniques to evaluate the impact of weather information on revisit intention utilizing reviews from TripAdvisor and online weather data. Information retrieval techniques (Doc2Vec) are applied on online reviews collected during the summer months between 2010-2019 from tourists that visited Cyprus. Reviews are labeled as "revisits" or "neutral" based on their textual content. The labelled reviews dataset is enhanced with weather information based on the reviews’ timestamp, such as temperature and humidity of tourists’ country of origin and Cyprus at the time of the visit to the hotel/destination. To account for the influence of hotel infrastructure and available services to deal with heatwaves (i.e., climate-controlled), the training dataset included hotel star rating as an additional parameter. An ensemble gradient boosting tree classifier is trained utilizing the compiled dataset to predict revisit intention. The classifier is evaluated against the area under the curve. To interpret the classifier’s inherent patterns, a popular machine learning interpretation technique is used, namely Shapley Additive Explanation (SHAP). Visualizations of the model using SHAP indicate that the heat index and weather difference between destination and country of origin influence revisit intention. Such preliminary insights are encouraging for further investigations with an end goal to develop a decision support system to assist destination managers during their target marketing campaigns.