Analysis of Visitor Review Data Using Lexicon Based, Support Vector Machine, Random Forest in Determining the Priority Scale of Building Labuan Bajo Tourism Objects

Abstract

Labuan Bajo tourist destination is one of the super priority tourist destinations in Indonesia. The importance of obtaining and analyzing tourists' reviews is to understand their preferences and views on the existing facilities and services. Therefore, this research is conducted to obtain and analyze visitor review data obtained from TripAdvisor and Google Maps. The methods used in analyzing these visitor reviews are Lexicon-Based for labeling, Support Vector Machine (SVM), and Random Forest for classification. The labeling results using the Lexicon-Based method showed 4187 positive reviews, 1796 negative reviews, and 1774 neutral reviews. The classification was performed using SMOTE (Synthetic Minority Over-sampling Technique) and without using SMOTE due to data imbalance. Results using SMOTE with SVM showed an accuracy of 0.89, precision of 0.95, recall of 0.85, and f1-measure of 0.90, with an ROC AUC value of 0.94, with Random Forest showed an accuracy of 0.87, precision of 0.91, recall of 0.86, and f1-measure of 0.88, with an ROC AUC value of 0.93. The determination of priority scale was done by obtaining the top 10 words and the number of sentiments related to development. The frequently occurring positive sentiment words were 'beautiful,' 'natural,' 'exotic,' 'scenic,' 'clean,' 'ancient,' 'amazed,' and 'historical.' The preservation of natural and historical assets must be maintained and continuously preserved.On the other hand, the frequently occurring negative words were 'expensive,' 'cost,' 'guide,' 'road,' 'garbage,' and 'hot.' Based on these words, the development of transportation and infrastructure is undoubtedly needed to enhance the attractiveness of Labuan Bajo as a tourist destination.