AZRAI, NABIL (2026) ANALISIS PENGARUH LSTM SEBAGAI EKSTRAKTOR FITUR TERHADAP KINERJA KLASIFIKASI SENTIMEN KUALITAS BBM MENGGUNAKAN FASTTEXT DAN SVM. S1 thesis, Universitas Mercu Buana Jakarta.
|
Text (HAL COVER)
01 Cover.pdf Download (638kB) | Preview |
|
![[img]](https://repository.mercubuana.ac.id/style/images/fileicons/text.png) |
Text (BAB I)
02 Bab 1.pdf Restricted to Registered users only Download (48kB) |
|
|
Text (BAB II)
03 Bab 2.pdf Restricted to Registered users only Download (274kB) |
|
|
Text (BAB III)
04 Bab 3.pdf Restricted to Registered users only Download (40kB) |
|
|
Text (BAB IV)
05 Bab 4.pdf Restricted to Registered users only Download (377kB) |
|
|
Text (BAB V)
06 Bab 5.pdf Restricted to Registered users only Download (40kB) |
|
|
Text (DAFTAR PUSTAKA)
07 Daftar Pustaka.pdf Restricted to Registered users only Download (163kB) |
|
|
Text (LAMPIRAN)
08 Lampiran.pdf Restricted to Registered users only Download (564kB) |
Abstract
Fuel quality in Indonesia has become one of the most widely discussed public issues, particularly on the X (formerly Twitter) platform. Opinions expressed by users contain positive, neutral, and negative sentiments, making sentiment analysis an effective approach to identify public perceptions regarding fuel quality. This study aimed to analyze the effect of using Long Short-Term Memory (LSTM) as a feature extractor on sentiment classification performance using Support Vector Machine (SVM). A quantitative experimental approach was employed by comparing two classification scenarios: FastText-SVM and FastText-LSTM-SVM. The dataset was collected through scraping X and consists of 3,931 tweets published between 2016 and May 19, 2026. The research process included data preprocessing, word embedding using FastText, feature extraction using LSTM in the second scenario, sentiment classification using SVM, and model evaluation using a confusion matrix based on accuracy, precision, recall, and F1-score. The experimental results showed that the FastText-SVM model achieved an accuracy of 76.90%, while the FastText-LSTM-SVM model achieved 75.53%. Although the use of LSTM did not improve the overall classification performance, it enhanced the model's ability to recognize the neutral sentiment class, with recall increasing from 0.40 to 0.48 and the F1-score improving from 0.47 to 0.50. These findings indicate that the use of LSTM as a feature extractor did not improve the overall classification performance on the dataset used in this study, but it contributed to better recognition of neutral sentiment. Kata kunci: Sentiment Analysis, FastText, Fuel Quality, Long Short-Term Memory (LSTM), Support Vector Machine (SVM). Kualitas bahan bakar minyak (BBM) di Indonesia menjadi salah satu isu yang banyak diperbincangkan masyarakat, khususnya melalui platform X (Twitter). Beragam opini yang disampaikan pengguna media sosial mengandung sentimen positif, netral, dan negatif sehingga diperlukan analisis sentimen untuk mengetahui persepsi masyarakat terhadap kualitas BBM. Penelitian ini bertujuan menganalisis pengaruh penggunaan Long Short-Term Memory (LSTM) sebagai ekstraktor fitur terhadap performa klasifikasi sentimen menggunakan Support Vector Machine (SVM). Penelitian dilakukan menggunakan pendekatan kuantitatif dengan metode eksperimen melalui perbandingan dua skenario, yaitu FastText-SVM dan FastTextLSTM-SVM. Data penelitian diperoleh melalui proses scraping pada platform X sebanyak 3.931 tweet yang dipublikasikan pada periode 2016 hingga 19 Mei 2026. Tahapan penelitian meliputi preprocessing, pembentukan word embedding menggunakan FastText, ekstraksi fitur menggunakan LSTM pada skenario kedua, proses klasifikasi menggunakan SVM, serta evaluasi menggunakan confusion matrix berdasarkan nilai accuracy, precision, recall, dan F1-score. Hasil penelitian menunjukkan bahwa skenario FastText-SVM memperoleh nilai accuracy sebesar 76,90%, sedangkan FastText-LSTM-SVM memperoleh nilai accuracy sebesar 75,53%. Meskipun tidak meningkatkan performa klasifikasi secara keseluruhan, penggunaan LSTM sebagai ekstraktor fitur mampu meningkatkan kemampuan model dalam mengenali kelas sentimen netral, yang ditunjukkan oleh peningkatan nilai recall dari 0,40 menjadi 0,48 serta F1-score dari 0,47 menjadi 0,50. Berdasarkan hasil tersebut, penggunaan LSTM sebagai ekstraktor fitur belum memberikan peningkatan performa klasifikasi secara keseluruhan pada dataset yang digunakan, namun memberikan kontribusi dalam meningkatkan pengenalan terhadap kelas sentimen netral. Kata kunci: Analisis Sentimen, FastText, Long Short-Term Memory (LSTM), Support Vector Machine (SVM), Kualitas BBM.
Actions (login required)
 |
View Item |
