ANALISIS SENTIMEN BERBASIS ASPEK PADA ULASAN PRODUK ELEKTRONIK LAPTOP DENGAN BIDIRECTIONAL LONG SHORT-TERM MEMORY (BI-LSTM)

Hikmawati Fajriah Ayu Wardana, - (2023) ANALISIS SENTIMEN BERBASIS ASPEK PADA ULASAN PRODUK ELEKTRONIK LAPTOP DENGAN BIDIRECTIONAL LONG SHORT-TERM MEMORY (BI-LSTM). S1 thesis, Universitas Pendidikan Indonesia.

Abstract

Natural Language Processing (NLP) merupakan bidang penelitian yang berfokus pada pemrosesan bahasa. Sentimen pelanggan pada ulasan produk laptop dapat dianalisis menggunakan teknik NLP. Dalam penelitian ini, analisis sentimen didasarkan pada topik tertentu sehingga disebut sebagai analisis sentimen berbasis aspek. Penelitian ini menggunakan metode deep learning untuk mengotomatisasi proses ekstraksi fitur pada data. Dataset untuk model diperoleh melalui web scraping ulasan laptop pada situs e-commerce di Indonesia dan diberi label dengan metode sequence labeling berformat BIO, menghasilkan 2 class yaitu aspek dan sentimen. Class aspek terdiri dari 7 label (B-BOD, I-BOD, B-POW, I-POW, B-KEY, I-KEY, dan O). Class sentimen terdiri dari 5 label (B-POS, I-POS, B-NEG, I-NEG, dan O). Model dibangun dengan algoritma Bidirectional Long Short-Term Memory (Bi-LSTM), menghasilkan nilai accuracy dan F1-score terbaik secara beurutan sebesar 0.865 dan 0.739 (model deteksi aspek), 0.846 dan 0.681 (model deteksi sentimen), serta 0.836 dan 0.611 (model deteksi aspek dan sentimen). Model ini diimplementasikan pada halaman web agar dapat menerima input dari pengguna, memberikan alat berharga untuk kepentingan bisnis dalam memahami feedback pelanggan.

Natural Language Processing (NLP) is a research field that focuses on language processing. Customer sentiments in laptop product reviews can be analyzed using NLP techniques. In this study, sentiment analysis is conducted based on specific topics, making it referred to as aspect-based sentiment analysis. Deep learning methods are employed to automate feature extraction from data. The dataset for the model is obtained through web scraping laptop reviews e-commerce site in Indonesia and labeled using the BIO format sequence labeling method, resulting in two classes: aspect and sentiment. The aspect class consists of 7 labels (B-BOD, I-BOD, B-POW, I-POW, B-KEY, I-KEY, and O). The sentiment class consists of 5 labels (B-POS, I-POS, B-NEG, I-NEG, and O). The model is built using the Bidirectional Long Short-Term Memory (Bi-LSTM) algorithm and achieves the best consecutive accuracy and F1-scores of 0.865 and 0.739 (aspect detection model), 0.846 and 0.681 (sentiment detection model), 0.836 and 0.611 (aspect and sentiment detection model). This model is implemented on a web page to receive user input, providing a valuable tool for businesses to deeply understand customer feedback.

	Text S_KOM_1903510_Title.pdf Download (377kB)
	Text S_KOM_1903510_Chapter1.pdf Download (290kB)
	Text S_KOM_1903510_Chapter2.pdf Restricted to Staf Perpustakaan Download (1MB)
	Text S_KOM_1903510_Chapter3.pdf Download (182kB)
	Text S_KOM_1903510_Chapter4.pdf Restricted to Staf Perpustakaan Download (1MB)
	Text S_KOM_1903510_Chapter5.pdf Download (119kB)

Item Type:	Thesis (S1)
Additional Information:	Google Scholar Dosen Pembimbing: https://scholar.google.com/citations?user=gdiW3PgAAAAJ&hl=en ID SINTA Dosen Pembimbing: Yudi Wibisono : 260167 Rani Megasari : 5992674
Uncontrolled Keywords:	Analisis Sentimen Berbasis Aspek, Pemrosesan Bahasa Alami, Ulasan Produk Elektronik Laptop, E-Commerce, Sequence Labeling, Deep Learning, Bi-LSTM Aspect-Based Sentiment Analysis, Natural Language Processing, Laptop Electronic Product Reviews, E-Commerce, Sequence Labeling, Deep Learning, Bi-LSTM
Subjects:	L Education > L Education (General) Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:	Fakultas Pendidikan Matematika dan Ilmu Pengetahuan Alam > Program Studi Ilmu Komputer
Depositing User:	Hikmawati Fajriah Ayu Wardana
Date Deposited:	04 Jan 2024 06:16
Last Modified:	04 Jan 2024 06:16
URI:	http://repository.upi.edu/id/eprint/113994

Actions (login required)

View Item