Istadewanti, Anak Agung Istri (2025) Analisis Sentimen Pada Desa Wisata Berkonsep Sustainable Tourism Berdasarkan Ulasan Pada Google Maps Menggunakan EMC- GCN Dan LSTM-CNN. Other thesis, Institut Teknologi Sepuluh Nopember.
![]() |
Text
5026211143-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (9MB) | Request a copy |
Abstract
Desa wisata merupakan kawasan yang menawarkan pengalaman khas pedesaan seperti keindahan alam, tradisi, dan elemen unik lainnya yang secara keseluruhan mampu menarik minat wisatawan. Untuk mendukung pengembangannya, analisis sentimen dapat digunakan untuk memahami persepsi wisatawan dari ulasan yang tersedia. Tugas akhir ini menganalisis ulasan berbahasa Indonesia dari enam desa wisata berkonsep sustainable tourism menurut Kemenparekraf, yaitu Desa Wisata Pujon Kidul, Pentingsari, Kete Kesu, Kampung Blekok, Umbul Ponggok, dan Penglipuran. Data diambil melalui Google Maps yang dibersihkan, lalu dianalisis pada dua tingkatan, dokumen dan aspek. Pada tingkat dokumen, model CNN-LSTM dengan IndoBERT embedding dan data seimbang menghasilkan performa terbaik dengan F1- score sebesar 92,88%. Meskipun begitu, model masih mengalami kesalahan klasifikasi pada ulasan yang menyampaikan opini atau emosi secara tersirat. Namun, secara keseluruhan model tetap menunjukkan performa yang stabil dengan tingkat kesalahan yang rendah. Sementara itu, pada tingkat aspek, model EMC-GCN dengan modifikasi fungsi “find_triplet” mencatat F1- score sebesar 67,61%. Melalui analisis hasil evaluasi, performa model EMC-GCN lebih optimal pada ulasan yang eksplisit dan terstruktur jelas, serta lebih stabil pada kategori dengan jumlah data yang melimpah. Evaluasi lanjutan pada tingkat aspek menunjukkan bahwa tidak semua prediksi yang berada diluar kategori exact match dapat dianggap salah, sebagian masih dapat diterima sebagai partial match. Bahkan pada kategori wrong prediction, beberapa hasil prediksi tetap mencerminkan isi ulasan meskipun tidak tercantum dalam label aktual, sehingga masih relevan untuk dianalisis lebih lanjut. Berdasarkan hasil ekstraksi triplet, aspek yang paling sering disebut meliputi “tempat”, “pemandangan”, “suasana”, “harga tiket masuk”, dan “warga sekitar” beserta ciri khas dari setiap desa untuk sentimen positif, serta “toilet”, “akses”, dan “tempat parkir” untuk sentimen negatif. Sementara itu, sentimen netral cenderung memberikan informasi terkait “desa wisata”, “harga tiket masuk” atau fasilitas, layanan, dan lokasi dari desa wisata tersebut.
=================================================================================================================================
Tourism village is an area that offers unique rural experiences such as natural beauty, traditions, and other distinctive elements that collectively attract tourists. To support its development, sentiment analysis can be used to understand tourists perceptions based on available reviews. This undergraduate thesis analyzes Indonesian-language reviews from six tourism villages with a sustainable tourism concept according to the Kemenparekraf, namely Pujon Kidul, Pentingsari, Kete Kesu, Kampung Blekok, Umbul Ponggok, and Penglipuran. The data were obtained from Google Maps, then preprocessed and analyzed at both the document level and the aspect level. At the document level, the CNN-LSTM model with IndoBERT embeddings and balanced data achieved the best performance, reaching an F1-score of 92,88%. Although some classification errors occurred in reviews that implicitly expressed opinions or emotions, the model overall demonstrated stable performance with a low error rate. Meanwhile, at the aspect level, the EMC-GCN model with a modified “find_triplet” function achieved an F1 score of 67,61%. Through analysis of the evaluation results, the EMC-GCN model performed better on explicit and clearly structured reviews and was more stable in categories with abundant data. Further evaluation at the aspect level shows that not all predictions outside the exact match category can be considered incorrect, some can still be accepted as partial matches. Even in the incorrect prediction category, some prediction results still reflect the content of the review even though they are not listed in the actual label, making them relevant for further analysis. Based on the triplet extraction results, the most frequently mentioned aspects include “place”, “scenery”, “atmosphere”, “entrance ticket prices”, and “local residents” along with the unique characteristics of each village for positive sentiment, as well as “toilets”, “access”, and “parking” for negative sentiment. Meanwhile, neutral sentiment tends to provide information related to “tourism villages”, “entrance ticket prices”, or facilities, services, and the location of the tourist village.
Item Type: | Thesis (Other) |
---|---|
Uncontrolled Keywords: | Analisis Sentimen Berbasis Aspek, Desa Wisata, Transfer Learning, Ulasan Google Maps, Aspect-Based Sentiment Analysis, Google Maps Review, Tourism Village, Transfer Learning |
Subjects: | G Geography. Anthropology. Recreation > G Geography (General) > G155 Tourism Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. T Technology > T Technology (General) > T57.5 Data Processing |
Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis |
Depositing User: | Anak Agung Istri Istadewanti |
Date Deposited: | 21 Jul 2025 03:02 |
Last Modified: | 21 Jul 2025 03:02 |
URI: | http://repository.its.ac.id/id/eprint/120184 |
Actions (login required)
![]() |
View Item |