Comparison of Feature Extraction Methods on Sentiment Analysis in Hotel Reviews

Abstract

The development of technology causes things that done through meet in person or coming to a place can now be done by viewing information through gadgets or websites. Nowadays, to find out information about a place that provides accommodation for a vacation or a business visit, it can be done by accessing social media to see reviews from visitors who have visited the place, example, a hotel. Reviews given by hotel visitors are seen as more credible than information obtained from advertisements but the problem is that there are many reviews circulating on social media and it takes a time to analyze them. This study aims to analyze hotel reviews using the sentiment analysis method with the Support Vector Machine (SVM) approach. Sentiment analysis can be used to analyze the opinions of a large number of hotel visitors where it usually focuses on opinions that positive, negative and neutral. Before being analyzed with the support vector machine algorithm, 3 feature extraction methods will be used, namely Bag Of Words, TF-IDF and improvement TF-IDF to get the value of each word weight. The selection of these three methods is carried out by considering the influence of the presence of the same word feature in each review. In this comparison method, TF-IDF was found to be the best feature extraction method with 71.75% accuracy, 78.66% precision, 71.91% recall and 70.08% f1-score. The results obtained indicate that there are influence of features of the word in the hotel review data.