LSTM-based Deep Learning Architecture of Tourist Review in Tripadvisor

Abstract

Information technology has become an essential component in various fields, one of which is in the world of tourism. This can be seen with the emergence of websites in the tourism sector, such as TripAdvisor. Bali, as a tourist destination that is well known to foreign countries, can use TripAdvisor as a place to promote tourist attractions in Bali, such as beaches. This study will conduct a sentiment analysis of tourist reviews on the five most favorite beaches in Bali, namely Double Six Beach, Seminyak, Nusa Dua, Delinking, and Cangue, on the TripAdvisor Website. The analysis results determine how tourists' opinions of the five beaches are using the Long Short-Term Memory (LSTM) architecture with sentiment classified in the form of positive and negative labels. This study uses a training and testing ratio of 80:20 with the undersampling method because it has the highest overall accuracy with Double Six Beach 83%, Seminyak 81%, Nusa Dua 84%, Kelingking 81%, and Canggu 84%. The prediction results obtained from the classification results are more dominant on the positive label. In addition to classifying for sentiment analysis, this study also measures the model’s performance created by calculating the value of precision, recall, F1-Score, macro average, and weight average for each beach classification result.