Sentiment Analysis on Social Media Using Long Short-Term Memory and Word2Vec Feature Expansion methods with Adam Optimization - Dalam bentuk pengganti sidang - Artikel Jurnal

SANABILA KHOIRUNNISA

Informasi Umum

Kode

24.04.758

Klasifikasi

001.64 - DATA PROCESSING

Jenis

Karya Ilmiah - Skripsi (S1) - Reference

Subjek

Data Science

Dilihat

14 kali

Informasi Lainnya

Abstraksi

<p>Twitter is one of Indonesia's most popular social media, so it has many users. The intensity of Twitter use can be used to carry out sentiment analysis related to topics being widely discussed, especially regarding the 2024 Indonesian presidential election. To understand public views, public opinion is divided from text data into positive and negative polarities to measure public sentiment. The classification model uses Long Short-Term Memory (LSTM) for feature extraction, utilizing TF-IDF. In addition, this model also combines Word2Vec based on the Indonews corpus, which contains 142,545 articles for feature expansion. This model is further optimized using the Adam optimization technique to improve accuracy. By using a dataset of 37,391 data, the results of this research obtained an accuracy score of 83.04% and an f1 score of 82.62%. This is an increase in accuracy of 11.22%; for the f1 score, it was a 10.92% increase from the baseline. This indicates that the classification model using Long Short-Term Memory (LSTM) with the application of TF-IDF as feature extraction, Word2Vec as feature expansion, and Adam optimization successfully produced optimal sentiment predictions regarding the 2024 Indonesian Presidential Election.</p>