Identification of Disney Heavy Consumers on Twitter

Analysis of consumers' identity, behavior, and opinions is possible from a collection of their digital activity data on Twitter (X) and other social media platforms. Feedback from “heavy consumers” is particularly important for companies and directly translates into revenue. However, manually identifying which Twitter users are such customers is time-consuming and costly. In this study, we identified heavy consumers of Tokyo Disney Resorts among Twitter users. We considered various complex characteristics of such heavy consumers: for example, their visitation frequency of the Resorts, the distance between their home and the Resorts, and their purchasing activities of Tokyo Disney products. We collected a total of 100,000 tweets posted by the heavy users of the Tokyo Disney Resorts and developed a BERT (bidirectional encoder representations from transformers) model to test our dataset. Our model scored more than 70% of accuracy in Precision for the text without emoji. Our model scored higher for the text without emoji (70% of accuracy in Precision), indicating that BERT processes pictogram-free text more easily. Thereafter, our model is also more accurate than ChatGPT in analyzing both types of text. As a result, the accuracy of our BERT model was higher than that of ChatGPT.