Feature Extraction for Improvement Text Classification of Spam YouTube Video Comment using Deep Learning

Feature Extraction for Improvement Text Classification of Spam YouTube Video Comment using Deep Learning

Improvement; BLSTM; CRF; Data Augmentation; Feature Extraction

The proposed algorithms are Bidirectional Long Short Term Memory (BiLSTM) and Conditional Random Fields (CRF) with
Data Augmentation Technique (DAT). DAT integrates spam YouTube video comments into the traditional TF-IDF algorithm
and generates a weighted word vector. The weighted word vector is fed into BiLSTM CRF to capture context information
effectively. The result of this study is a new classification model to spam YouTube comment videos and increase the
computational value of its performance. This research conducted two experiments: the first using BiLSTM CRF without DAT
and the second using BiLSTM CRF with DAT. The experimental results state that the evaluation score using BiLSTM CRF with
DAT shows outstanding performance in text classification, especially in spam YouTube video comment texts, with accuracy =
83.3%, precision = 83.6%, recall = 83.3%, and F-measure = 83.3%. So the combination of the BiLSTM-CRF method and the
Data Augmentation Technique is very precise, so it can be used to increase the accuracy of classification texts for spam
YouTube video comments

Jasmir Jasmir, Willy Riyadi, Pareza Alam Jusia

http://jurnal.iaii.or.id

Professional Organization Ikatan Ahli Informatika Indonesia (IAII)/Indonesian Informatics Experts Association

December 2023

Sri Wahyuni

ISSN Media Electronic: 2580-0760

PDF

English

Text