Comparison of The Accuracy of K-Nearest Neighbor and Roberta Algorithm in Analysis of Sentiment on Miawaug Youtube Channel Comments

This study aims to evaluate the accuracy of two algorithms, K-Nearest Neighbor (KNN) and Robustly Optimized BERT Approach (RoBERTa), in analyzing sentiment within comments on MiawAug’s YouTube channel. Sentiment analysis was conducted on two sentiment categories: binary classification (positive and...

Full description

Saved in:
Bibliographic Details
Main Authors: Fachrudin Okta Rahmawan, Hanafi, Windha Mega Pradnya Dhuita
Format: Article
Language:English
Published: Center for Research and Community Service, Institut Informatika Indonesia Surabaya 2025-03-01
Series:Teknika
Subjects:
Online Access:https://ejournal.ikado.ac.id/index.php/teknika/article/view/1117
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study aims to evaluate the accuracy of two algorithms, K-Nearest Neighbor (KNN) and Robustly Optimized BERT Approach (RoBERTa), in analyzing sentiment within comments on MiawAug’s YouTube channel. Sentiment analysis was conducted on two sentiment categories: binary classification (positive and negative) and multi-class classification (positive, neutral, and negative). Using KNN, the binary classification yielded an accuracy of 86.12%, F1-score of 87.44%, recall of 96.64%, and precision of 79.89%. In contrast, the multi-class classification achieved 98.21% accuracy, F1-score, and recall with a precision of 98.23%. However, the RoBERTa model outperformed KNN, achieving 93.89% accuracy, 93.88% F1-score, 94.59% recall, and 93.22% precision in binary classification. For multi-class classification, RoBERTa further excelled, attaining 99.21% across accuracy, F1-score, recall, and precision. These findings demonstrate that RoBERTa surpasses KNN in sentiment analysis, especially in multi-class contexts, indicating its greater robustness for this application.
ISSN:2549-8037
2549-8045