Tourist Recommendations Based on Online Reviews with Sentiment Analysis using BERT in Central Java Tourism

Indonesia offers a variety of tourist attractions that attract many visitors, both local and international tourists. One of the most popular destinations is Central Java. The development of digital tourism services allows the availability of abundant data on various digital platforms. This study analyzes online reviews in the form of text and numerical data that have a major influence on consumer purchasing decisions. This study aims to propose a tourist destination recommendation model in Central Java that can effectively predict consumer assessment behavior by adding sentiment analysis variables using Bidirectional Encoder Representations for Transformers (BERT). The dataset used in this study is from the Tripadvisor website, offering comprehensive insights into various tourist attractions and reviews from tourists who have explored various locations in Central Java. The methodological design used by the author is to collect 2847 reviews from 2004 consumers on TripAdvisor. Using this data, the author produces a recommendation model for the top-7 tourist recommendations. This study found that the proposed recommendation model showed good performance; the model evaluation showed an MAE of 0.11620 and RMSE of 0.12143. In the future, the results of this study can be developed to reduce consumer exploration costs and support effective purchasing decisions.