The Accuracy Comparison of Social Media Sentiment Analysis Using Lexicon Based and Support Vector Machine on Souvenir Recommendations

Wilis, Kaswidjanti and Hidayatulah, Himawan and Parasian, Silitonga (2020) The Accuracy Comparison of Social Media Sentiment Analysis Using Lexicon Based and Support Vector Machine on Souvenir Recommendations. TEST Engineering & Management, 83. pp. 3953-3961. ISSN 0193-4120

[img]
Preview
Text
3. the accuracy comparison.pdf

Download (431kB) | Preview
Official URL: https://testmagzine.biz/index.php/testmagzine

Abstract

Social media is one source of information that can be obtained by organizations / vendors or consumers. One of the information obtained from social media is the opinion of social media users on a matter. These opinions can be positive, negative or neutral. Sentiment analysis can be used to assess the sentiments of the opinions expressed by social media users. Sentiment analysis method used is a machine learning algorithm that can help the classification process. In this study the comparison of the accuracy of opinion sentiment analysis on the recommendation of favorite souvenirs in the Yogyakarta area using the Lexicon Based method and Support Vector Machine. The processed data is Twitter and Instagram social media sentiment data. The training data used were 1000 sentiment data, while the test data for the testing process were 50 sentiment data. The test results obtained the greatest accuracy is using a Lexicon Based 87.78%, then the greatest precision results are using a Lexicon Based of 94.23%, while for the greatest recall results are using a Support Vector Machine of 100%.

Item Type: Article
Uncontrolled Keywords: ouvenir Recommendations, Sentiment Analysis, Social Media, Twitter, Instagram, Lexicon Based, Support Vector Machine
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Depositing User: ST,MM,MEng HIDAYATULAH HIMAWAN
Date Deposited: 27 Jul 2020 10:20
Last Modified: 27 Jul 2020 10:22
URI: http://eprints.upnyk.ac.id/id/eprint/23088

Actions (login required)

View Item View Item