Modelling the e-WOM Affective Image of Coastal Tourism Destinations

Tourism Destination Image (TDI) is a central construct in tourism research, with the affective dimension recognised as a key driver of tourists’ emotions, evaluations and behavioural intentions. However, empirical studies examining affective image through electronic word-of-mouth (e-WOM) remain limited, particularly when using large-scale, naturally occurring data. This study addresses this gap by modelling the e-WOM affective and overall image of Brazilian coastal tourism destinations through the integration of microblog data, supervised machine learning (ML), confirmatory factor analysis (CFA) and covariance-based structural equation modelling (CB-SEM). A dataset of 82,678 tweets from 16 coastal destinations was classified using an ordinal supervised ML approach trained on human-annotated data. The models achieved fair to substantial agreement with human judgements, demonstrating their ability to capture nuanced emotional expressions. CFA validated the measurement structure of the affective image construct, confirming high internal consistency and convergent validity. CB-SEM results further showed a significant and positive effect of e-WOM affective image on overall image. The study advances TDI theory by empirically validating affective images derived from user-generated content and by demonstrating the methodological value of combining ML and psychometric modelling. The findings also provide Destination Management Organisations (DMOs) with real-time analytical tools to monitor tourists’ emotions and strengthen coastal tourism destination branding.