Search for collections on Undip Repository

PENERAPAN BERT DAN LSTM DENGAN ATTENTION MECHANISM UNTUK PENENTUAN INDEKS KEPUASAN HIDUP MASYARAKAT BERBASIS ANALISIS SENTIMEN

WICAKSANA, Hilman Singgih and Kusumaningrum, Retno and Gernowo, Rahmat (2024) PENERAPAN BERT DAN LSTM DENGAN ATTENTION MECHANISM UNTUK PENENTUAN INDEKS KEPUASAN HIDUP MASYARAKAT BERBASIS ANALISIS SENTIMEN. Masters thesis, UNIVERSITAS DIPONEGORO.

[img] Text
Cover.pdf

Download (3MB)
[img] Text
Bab I.pdf

Download (1MB)
[img] Text
Bab II.pdf

Download (2MB)
[img] Text
Bab III.pdf
Restricted to Repository staff only

Download (2MB)
[img] Text
Bab IV.pdf
Restricted to Repository staff only

Download (3MB)
[img] Text
Bab V.pdf

Download (1MB)
[img] Text
Daftar Pustaka.pdf

Download (1MB)
[img] Text
Lampiran.pdf
Restricted to Repository staff only

Download (2MB)

Abstract

Di era digital saat ini, mengevaluasi kualitas hidup dan indeks kepuasan hidup masyarakat sangat erat kaitannya dengan ekspresi dan opini mereka di media sosial X (Twitter). Pengukuran kesejahteraan manusia tidak sebatas pada aspek moneter saja, melainkan lebih menitikberatkan pada kesejahteraan subjektif. Oleh karena itu pada penelitian ini diperlukan pendekatan berbasis analisis sentimen agar dapat membantu untuk mengevaluasi persepsi masyarakat terhadap indikator kepuasan hidup secara komprehensif. Pendekatan Aspect-Based Sentiment Analysis (ABSA) secara efektif mengidentifikasi sentimen pada aspek-aspek yang telah ditentukan. Penelitian sebelumnya telah menggunakan metode Word-to-Vector (Word2Vec) dan Long Short-Term Memory (LSTM) dengan atau tanpa Attention Mechanism (AM) untuk menyelesaikan kasus ABSA. Namun, masalah pada penelitian sebelumnya adalah Word2Vec memiliki kelemahan yaitu tidak dapat menangani konteks kata dalam sebuah kalimat. Bidirectional Encoder Representations from Transformers (BERT) mampu menangani permasalahan tersebut dengan melakukan pelatihan secara dua arah. Oleh karena itu penelitian ini menerapkan model BERT-LSTM-AM untuk penentuan indeks kepuasan hidup masyarakat. Bayesian Optimization sebagai teknik hyperparameter tuning digunakan untuk menemukan kombinasi parameter terbaik selama proses pelatihan. Hasil dari penelitian ini berupa sistem prediksi berbasis web yang menunjukkan bahwa masyarakat Jawa Tengah memiliki tingkat kepuasan hidup sekitar 66,4% dengan kategori kepuasan yaitu “Cukup Puas”.
Kata kunci: Indeks Kepuasan Hidup, Aspect-Based Sentiment Analysis, BERT, LSTM, X (Twitter)

In the current digital era, evaluating people’s quality of life and life satisfaction index is closely related to their expressions and opinions on X (Twitter) social media. The measurement of human well-being is not limited to the monetary aspect, but focuses more on subjective well-being. Therefore, this study requires a sentiment analysis-based approach to help evaluate people's perceptions of life satisfaction indicators comprehensively. The Aspect-Based Sentiment Analysis (ABSA) approach effectively identifies sentiment on predetermined aspects. Previous study has used Word-to-Vector (Word2Vec) and Long Short-Term Memory (LSTM) methods with or without Attention Mechanism (AM) to solve ABSA cases. However, the problem with previous study is that Word2Vec has the disadvantage of not being able to handle the context of words in a sentence. Bidirectional Encoder Representations from Transformers (BERT) is able to handle these problems by training in two directions. Therefore, this study applies BERT-LSTM-AM model to determine the community life satisfaction index. Bayesian Optimisation as a hyperparameter tuning technique is used to find the best combination of parameters during the training process. This study shows that BERT-LSTM-AM outperforms the Word2Vec-LSTM-AM model in predicting aspects and sentiments. In addition, it was found that BERT is an excellent embedding technique in representing words in a sentence. The result of this research is a web-based prediction system which shows that the people of Central Java have a life satisfaction level of around 66.4% with a satisfaction category of “Quite Satisfied”.
Keywords: Life Satisfaction Index, Aspect-Based Sentiment Analysis, BERT, LSTM, X (Twitter)

Item Type: Thesis (Masters)
Uncontrolled Keywords: Indeks Kepuasan Hidup, Aspect-Based Sentiment Analysis, BERT, LSTM, X (Twitter)
Subjects: Sciences and Mathemathic
Divisions: Postgraduate Program > Master Program in Information System
Depositing User: ekana listianawati
Date Deposited: 06 May 2024 07:53
Last Modified: 06 May 2024 07:53
URI: https://eprints2.undip.ac.id/id/eprint/22807

Actions (login required)

View Item View Item