Implementation of SVM, k-NN, and DT for Toxicity and Sentiment Classification of AWA Vlog Content in Wasur National Park

This study delves into the response of viewers to video content focusing on Wasur National Park in Papua, Indonesia, with a particular emphasis on its implications for livelihood and ecology. The increasing popularity of online platforms such as YouTube has provided a medium for content creators to...

Full description

Saved in:
Bibliographic Details
Published inJURNAL MEDIA INFORMATIKA BUDIDARMA Vol. 8; no. 2; p. 922
Main Author Singgalen, Yerik Afrianto
Format Journal Article
LanguageEnglish
Published 30.04.2024
Online AccessGet full text

Cover

Loading…
More Information
Summary:This study delves into the response of viewers to video content focusing on Wasur National Park in Papua, Indonesia, with a particular emphasis on its implications for livelihood and ecology. The increasing popularity of online platforms such as YouTube has provided a medium for content creators to showcase natural landscapes and cultural heritage, potentially influencing viewers' perceptions and behaviors toward conservation efforts. Employing the CRISP-DM (Cross-Industry Standard Process for Data Mining) framework, this research systematically analyzes a specific video from the AWA channel, known for its documentaries on environmental and cultural topics. The methodology involves sentiment analysis to gauge viewers' emotional responses, toxicity assessment to identify harmful content, and thematic coding to categorize comments based on recurring themes. The analysis reveals that viewers engage with the content positively, expressing appreciation for the video's educational and visually compelling nature. Moreover, the study identifies various dimensions of toxicity within the dataset, including Toxicity (0.05364), Severe Toxicity (0.00629), Identity Attack (0.02250), Insult (0.03534), Profanity (0.03589), and Threat (0.01280). Furthermore, the performance of the Support Vector Machine (SVM) with Synthetic Minority Over-sampling Technique (SMOTE) is highlighted, demonstrating its effectiveness in classifying sentiment with an accuracy of 93.86%, precision of 100.00%, recall of 87.73%, f-measure of 93.44%, and an Area Under the Curve (AUC) value of 1.000. This research underscores the significance of balanced media portrayals in fostering positive attitudes toward environmental conservation and cultural preservation.
ISSN:2614-5278
2548-8368
DOI:10.30865/mib.v8i2.7434