Comparison of Sentiment Analysis Methods Based on Accuracy Value Case Study: Twitter Mentions of Academic Article

Jurnal RESTI: Rekayasa Sistem dan Teknologi Informasi

article

Muhamad Fahmi Fakhrezi¹  Adian Fatchur Rochim¹  Dinar Mutiara Kusomo Nugraheni¹ 

[1] Universitas Diponegoro

关键词: Sentiment Analysis;    Decision Tree;    K-Nearest Neighbors;    Naïve Bayes;    Random Forest;   

DOI  :  10.29207/resti.v7i1.4767

来源: Ikatan Ahli Indormatika Indonesia

【 摘 要 】

The assessment of academic articles is based on the number of citations, but the number only is not enough. So now there is Altmetric which can measure the impact of academic articles from the number of citations and using social media, usually Twitter. Still, the number of mentions on Twitter is not enough because the expressions of the sentences vary. Mentions must be classified according to neutral, positive, and negative criteria. Sentiment analysis is performed on tweets to measure social media volume and attention related to research findings from academic articles. There are many sentiment analysis methods, so this study aims to compare sentiment analysis methods using Decision Tree, K-NN, Naïve Bayes, and Random Forest to get the most suitable methods. The evaluation method in this study uses the Confusion Matrix by searching for Accuracy, Precision, and Recall values. The results show that the most suitable sentiment analysis method is Naïve Bayes by obtaining the highest classification suitability value of the other methods, which has an actual positive sentiment value of neutral 2056, positive 1200, and negative 1292. In addition, Naïve Bayes gets the highest accuracy score of 95, 45%.

【 授权许可】

Unknown