Design of a Tourism Recommendation System Based on Sentiment Analysis with Lexicon LSTM

Abstract

The growing tourism sector encourages the development of a system that can assist tourists in finding the best tourist attractions. The system that is being popularly developed in tourism is the recommendation system. It provides recommendations for tourist attractions based on several parameters. One of the parameters that can help in developing a recommender system is tourist satisfaction. Tourist satisfaction can be known by sentiment analysis on reviews on social media or tourism sites. One way of sentiment analysis is using machine learning. However, machine learning requires quite a lot of data labeled for learning. This process takes a lot of time, especially for a large amount of data, so the idea arose to use a lexicon dictionary. In this study, we tested two Indonesian datasets with four classifiers SVM, Random Forest, KNN, and LSTM. Its classifiers will combine with lexicon corpus, BOW (Bag of Words), and TFIDF (Term Frequency Inverse Document Frequency) to determine the best combination method. The experiments found that the scenario lexicon corpus, The LSTM achieves the best performance for all datasets with each accuracy is 90.93% and 76.98%. Meanwhile, in the scenario of lexicon corpus-BOW and lexicon corpus-TFIDF, the SVM method is the best on all datasets. In the Online Lecture dataset, it gets S7.40% and S7.36% of accuracy. On IDSA Dataset, lexicon corpus-BOW, and lexicon corpus-TFIDF, it gets 73.71% and 72.97% of accuracy.