THE DEVELOPMENT OF TOURISM RECOMMENDATION SYSTEM WITH SENTIMENT ANALYSIS AND PREFERENCE EXTRACTION BASED ON TOURIST ATTRACTION REVIEWS

• Main Author: Lubihana, Elan
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/72089
• Institution: Institut Teknologi Bandung

One of the ways to revive the tourism industry is by strengthening tourism promotion through exhibition, event, or tourism recommendations based on tourists’ preferences. Several studies have been accomplished by building tourism recommendation systems to reach this opportunity. However, the solutions offered in those studies could not solve a few constraints regarding data availability, particularly the usage of social media data as the ground aspect for tourism recommendations. Meanwhile, some tourists do not own social media accounts and some others are inactive users. Consequently, a new approach is essential for tourism recommendations despite limited data availability. This research was carried out to offer a solution by extracting positive sentiments from Google Maps to determine the characteristics of tourist destinations in Bandung Raya using lexicon corpus. The satisfaction scores reveal of 45,58 % positive, 10,50% neutral, and 43,91% negative. Furthermore, the sentiment classification indicates that Support Vector Machine (SVM)- Bag of Words (BoW) and SVM- Term Frequency-Inverse Document Frequency (TF-IDF) achieve the better average accuracy values than Long Short-Term Memory (LSTM). Besides, K-Means method is applied, and it produces two significant groups of tourist attractions according to their similar characteristics. Each group contains 74 and 42 members of tourist attractions. In addition, the recommendation system gets higher than 0,5 precision for four or more recommendations.